Add files using upload-large-folder tool

approach/ovod/APE/ape/data/datasets/__init__.py

@@ -0,0 +1,19 @@
+from . import d_cube as _d_cube
+from . import flickr30k as _flickr30k
+from . import gqa as _gqa
+from . import grit as _grit
+from . import lvis_coco as _lvis_coco
+from . import lvis_coco_panoptic as _lvis_coco_panoptic
+from . import objects365 as _objects365
+from . import odinw_instance as _odinw_instance
+from . import oid as _oid
+from . import pascal_voc_external as _pascal_voc_external
+from . import phrasecut as _phrasecut
+from . import refcoco as _refcoco
+from . import register_bdd100k_panoseg as _register_bdd100k_panoseg
+from . import register_bdd100k_semseg as _register_bdd100k_semseg
+from . import register_pascal_context as _register_pascal_context
+from . import register_voc_seg as _register_voc_seg
+from . import sa1b as _sa1b
+from . import seginw_instance as _seginw_instance
+from . import visualgenome as _visualgenome

approach/ovod/APE/ape/data/datasets/coco.py

@@ -0,0 +1,383 @@
+import contextlib
+import io
+import logging
+import os
+
+import pycocotools.mask as mask_util
+
+from detectron2.data import DatasetCatalog, MetadataCatalog
+from detectron2.structures import BoxMode
+from detectron2.utils.file_io import PathManager
+from fvcore.common.timer import Timer
+
+"""
+This file contains functions to parse COCO-format annotations into dicts in "Detectron2 format".
+"""
+
+
+logger = logging.getLogger(__name__)
+
+__all__ = ["custom_load_coco_json", "custom_register_coco_instances"]
+
+
+def custom_load_coco_json(json_file, image_root, dataset_name=None, extra_annotation_keys=None):
+    """
+    Load a json file with COCO's instances annotation format.
+    Currently supports instance detection, instance segmentation,
+    and person keypoints annotations.
+
+    Args:
+        json_file (str): full path to the json file in COCO instances annotation format.
+        image_root (str or path-like): the directory where the images in this json file exists.
+        dataset_name (str or None): the name of the dataset (e.g., coco_2017_train).
+            When provided, this function will also do the following:
+
+            * Put "thing_classes" into the metadata associated with this dataset.
+            * Map the category ids into a contiguous range (needed by standard dataset format),
+              and add "thing_dataset_id_to_contiguous_id" to the metadata associated
+              with this dataset.
+
+            This option should usually be provided, unless users need to load
+            the original json content and apply more processing manually.
+        extra_annotation_keys (list[str]): list of per-annotation keys that should also be
+            loaded into the dataset dict (besides "iscrowd", "bbox", "keypoints",
+            "category_id", "segmentation"). The values for these keys will be returned as-is.
+            For example, the densepose annotations are loaded in this way.
+
+    Returns:
+        list[dict]: a list of dicts in Detectron2 standard dataset dicts format (See
+        `Using Custom Datasets </tutorials/datasets.html>`_ ) when `dataset_name` is not None.
+        If `dataset_name` is None, the returned `category_ids` may be
+        incontiguous and may not conform to the Detectron2 standard format.
+
+    Notes:
+        1. This function does not read the image files.
+           The results do not have the "image" field.
+    """
+    from pycocotools.coco import COCO
+
+    timer = Timer()
+    json_file = PathManager.get_local_path(json_file)
+    with contextlib.redirect_stdout(io.StringIO()):
+        coco_api = COCO(json_file)
+    if timer.seconds() > 1:
+        logger.info("Loading {} takes {:.2f} seconds.".format(json_file, timer.seconds()))
+
+    id_map = None
+    if dataset_name is not None:
+        meta = MetadataCatalog.get(dataset_name)
+        cat_ids = sorted(coco_api.getCatIds())
+        cats = coco_api.loadCats(cat_ids)
+        # The categories in a custom json file may not be sorted.
+        thing_classes = [c["name"] for c in sorted(cats, key=lambda x: x["id"])]
+        meta.thing_classes = thing_classes
+
+        # In COCO, certain category ids are artificially removed,
+        # and by convention they are always ignored.
+        # We deal with COCO's id issue and translate
+        # the category ids to contiguous ids in [0, 80).
+
+        # It works by looking at the "categories" field in the json, therefore
+        # if users' own json also have incontiguous ids, we'll
+        # apply this mapping as well but print a warning.
+        if not (min(cat_ids) == 1 and max(cat_ids) == len(cat_ids)):
+            if "coco" not in dataset_name:
+                logger.warning(
+                    """
+Category ids in annotations are not in [1, #categories]! We'll apply a mapping for you.
+"""
+                )
+        id_map = {v: i for i, v in enumerate(cat_ids)}
+        meta.thing_dataset_id_to_contiguous_id = id_map
+
+        cat_ids = cat_ids + list(range(max(cat_ids) + 1, 100000))
+        id_map = {v: i for i, v in enumerate(cat_ids)}
+
+    # sort indices for reproducible results
+    img_ids = sorted(coco_api.imgs.keys())
+    # imgs is a list of dicts, each looks something like:
+    # {'license': 4,
+    #  'url': 'http://farm6.staticflickr.com/5454/9413846304_881d5e5c3b_z.jpg',
+    #  'file_name': 'COCO_val2014_000000001268.jpg',
+    #  'height': 427,
+    #  'width': 640,
+    #  'date_captured': '2013-11-17 05:57:24',
+    #  'id': 1268}
+    imgs = coco_api.loadImgs(img_ids)
+    # anns is a list[list[dict]], where each dict is an annotation
+    # record for an object. The inner list enumerates the objects in an image
+    # and the outer list enumerates over images. Example of anns[0]:
+    # [{'segmentation': [[192.81,
+    #     247.09,
+    #     ...
+    #     219.03,
+    #     249.06]],
+    #   'area': 1035.749,
+    #   'iscrowd': 0,
+    #   'image_id': 1268,
+    #   'bbox': [192.81, 224.8, 74.73, 33.43],
+    #   'category_id': 16,
+    #   'id': 42986},
+    #  ...]
+    anns = [coco_api.imgToAnns[img_id] for img_id in img_ids]
+    total_num_valid_anns = sum([len(x) for x in anns])
+    total_num_anns = len(coco_api.anns)
+    if total_num_valid_anns < total_num_anns:
+        logger.warning(
+            f"{json_file} contains {total_num_anns} annotations, but only "
+            f"{total_num_valid_anns} of them match to images in the file."
+        )
+
+    if "minival" not in json_file:
+        # The popular valminusminival & minival annotations for COCO2014 contain this bug.
+        # However the ratio of buggy annotations there is tiny and does not affect accuracy.
+        # Therefore we explicitly white-list them.
+        ann_ids = [ann["id"] for anns_per_image in anns for ann in anns_per_image]
+        assert len(set(ann_ids)) == len(ann_ids), "Annotation ids in '{}' are not unique!".format(
+            json_file
+        )
+
+    imgs_anns = list(zip(imgs, anns))
+    logger.info("Loaded {} images in COCO format from {}".format(len(imgs_anns), json_file))
+
+    dataset_dicts = []
+
+    ann_keys = ["iscrowd", "bbox", "keypoints", "category_id"] + (extra_annotation_keys or [])
+
+    ann_keys += ["phrase", "isobject"]
+
+    num_instances_without_valid_segmentation = 0
+
+    for (img_dict, anno_dict_list) in imgs_anns:
+        record = {}
+        record["file_name"] = os.path.join(image_root, img_dict["file_name"])
+        record["height"] = img_dict["height"]
+        record["width"] = img_dict["width"]
+        image_id = record["image_id"] = img_dict["id"]
+        if "neg_category_ids" in img_dict:
+            record["neg_category_ids"] = [id_map[x] for x in img_dict["neg_category_ids"]]
+
+        objs = []
+        for anno in anno_dict_list:
+            # Check that the image_id in this annotation is the same as
+            # the image_id we're looking at.
+            # This fails only when the data parsing logic or the annotation file is buggy.
+
+            # The original COCO valminusminival2014 & minival2014 annotation files
+            # actually contains bugs that, together with certain ways of using COCO API,
+            # can trigger this assertion.
+            assert anno["image_id"] == image_id
+
+            assert anno.get("ignore", 0) == 0, '"ignore" in COCO json file is not supported.'
+
+            obj = {key: anno[key] for key in ann_keys if key in anno}
+            if "bbox" in obj and len(obj["bbox"]) == 0:
+                raise ValueError(
+                    f"One annotation of image {image_id} contains empty 'bbox' value! "
+                    "This json does not have valid COCO format."
+                )
+
+            segm = anno.get("segmentation", None)
+            if segm:  # either list[list[float]] or dict(RLE)
+                if isinstance(segm, dict):
+                    if isinstance(segm["counts"], list):
+                        # convert to compressed RLE
+                        segm = mask_util.frPyObjects(segm, *segm["size"])
+                else:
+                    # filter out invalid polygons (< 3 points)
+                    segm = [poly for poly in segm if len(poly) % 2 == 0 and len(poly) >= 6]
+                    if len(segm) == 0:
+                        num_instances_without_valid_segmentation += 1
+                        continue  # ignore this instance
+                obj["segmentation"] = segm
+
+            keypts = anno.get("keypoints", None)
+            if keypts:  # list[int]
+                for idx, v in enumerate(keypts):
+                    if idx % 3 != 2:
+                        # COCO's segmentation coordinates are floating points in [0, H or W],
+                        # but keypoint coordinates are integers in [0, H-1 or W-1]
+                        # Therefore we assume the coordinates are "pixel indices" and
+                        # add 0.5 to convert to floating point coordinates.
+                        keypts[idx] = v + 0.5
+                obj["keypoints"] = keypts
+
+            # phrase = anno.get("phrase", None)
+            # if phrase:
+            #     obj["phrase"] = phrase
+
+            # isobject = anno.get("isobject", None)
+            # if isobject:
+            #     obj["isobject"] = isobject
+
+            obj["bbox_mode"] = BoxMode.XYWH_ABS
+            if id_map:
+                annotation_category_id = obj["category_id"]
+                try:
+                    obj["category_id"] = id_map[annotation_category_id]
+                except KeyError as e:
+                    raise KeyError(
+                        f"Encountered category_id={annotation_category_id} "
+                        "but this id does not exist in 'categories' of the json file."
+                    ) from e
+            objs.append(obj)
+        record["annotations"] = objs
+        dataset_dicts.append(record)
+
+    if num_instances_without_valid_segmentation > 0:
+        logger.warning(
+            "Filtered out {} instances without valid segmentation. ".format(
+                num_instances_without_valid_segmentation
+            )
+            + "There might be issues in your dataset generation process.  Please "
+            "check https://detectron2.readthedocs.io/en/latest/tutorials/datasets.html carefully"
+        )
+    return dataset_dicts
+
+
+def custom_load_sem_seg(gt_root, image_root, gt_ext="png", image_ext="jpg"):
+    """
+    Load semantic segmentation datasets. All files under "gt_root" with "gt_ext" extension are
+    treated as ground truth annotations and all files under "image_root" with "image_ext" extension
+    as input images. Ground truth and input images are matched using file paths relative to
+    "gt_root" and "image_root" respectively without taking into account file extensions.
+    This works for COCO as well as some other datasets.
+
+    Args:
+        gt_root (str): full path to ground truth semantic segmentation files. Semantic segmentation
+            annotations are stored as images with integer values in pixels that represent
+            corresponding semantic labels.
+        image_root (str): the directory where the input images are.
+        gt_ext (str): file extension for ground truth annotations.
+        image_ext (str): file extension for input images.
+
+    Returns:
+        list[dict]:
+            a list of dicts in detectron2 standard format without instance-level
+            annotation.
+
+    Notes:
+        1. This function does not read the image and ground truth files.
+           The results do not have the "image" and "sem_seg" fields.
+    """
+
+    # We match input images with ground truth based on their relative filepaths (without file
+    # extensions) starting from 'image_root' and 'gt_root' respectively.
+    def file2id(folder_path, file_path):
+        # extract relative path starting from `folder_path`
+        image_id = os.path.normpath(os.path.relpath(file_path, start=folder_path))
+        # remove file extension
+        image_id = os.path.splitext(image_id)[0]
+        return image_id
+
+    input_files = sorted(
+        (os.path.join(image_root, f) for f in PathManager.ls(image_root) if f.endswith(image_ext)),
+        key=lambda file_path: file2id(image_root, file_path),
+    )
+    gt_files = sorted(
+        (os.path.join(gt_root, f) for f in PathManager.ls(gt_root) if f.endswith(gt_ext)),
+        key=lambda file_path: file2id(gt_root, file_path),
+    )
+
+    assert len(gt_files) > 0, "No annotations found in {}.".format(gt_root)
+
+    # Use the intersection, so that val2017_100 annotations can run smoothly with val2017 images
+    if len(input_files) != len(gt_files):
+        logger.warn(
+            "Directory {} and {} has {} and {} files, respectively.".format(
+                image_root, gt_root, len(input_files), len(gt_files)
+            )
+        )
+        input_basenames = [os.path.basename(f)[: -len(image_ext)] for f in input_files]
+        gt_basenames = [os.path.basename(f)[: -len(gt_ext)] for f in gt_files]
+        intersect = list(set(input_basenames) & set(gt_basenames))
+        # sort, otherwise each worker may obtain a list[dict] in different order
+        intersect = sorted(intersect)
+        logger.warn("Will use their intersection of {} files.".format(len(intersect)))
+        input_files = [os.path.join(image_root, f + image_ext) for f in intersect]
+        gt_files = [os.path.join(gt_root, f + gt_ext) for f in intersect]
+
+    logger.info(
+        "Loaded {} images with semantic segmentation from {}".format(len(input_files), image_root)
+    )
+
+    dataset_dicts = []
+    for (img_path, gt_path) in zip(input_files, gt_files):
+        record = {}
+        record["file_name"] = img_path
+        record["sem_seg_file_name"] = gt_path
+        dataset_dicts.append(record)
+
+    return dataset_dicts
+
+
+def custom_load_sem_seg_list(gt_root, image_root, gt_ext="png", image_ext="jpg"):
+    if isinstance(image_root, list):
+        image_roots = image_root
+    else:
+        image_roots = [image_root]
+    if isinstance(gt_root, list):
+        gt_roots = gt_root
+    else:
+        gt_roots = [gt_root]
+
+    dataset_dicts = []
+    for gt_root, image_root in zip(gt_roots, image_roots):
+        dataset_dicts.extend(custom_load_sem_seg(gt_root, image_root, gt_ext, image_ext))
+
+    if len(image_roots) > 1:
+        logger.info(
+            "Loaded {} images with semantic segmentation from {}".format(
+                len(dataset_dicts), image_roots
+            )
+        )
+
+    return dataset_dicts
+
+
+def custom_register_coco_instances(name, metadata, json_file, image_root):
+    """
+    Register a dataset in COCO's json annotation format for
+    instance detection, instance segmentation and keypoint detection.
+    (i.e., Type 1 and 2 in http://cocodataset.org/#format-data.
+    `instances*.json` and `person_keypoints*.json` in the dataset).
+
+    This is an example of how to register a new dataset.
+    You can do something similar to this function, to register new datasets.
+
+    Args:
+        name (str): the name that identifies a dataset, e.g. "coco_2014_train".
+        metadata (dict): extra metadata associated with this dataset.  You can
+            leave it as an empty dict.
+        json_file (str): path to the json instance annotation file.
+        image_root (str or path-like): directory which contains all the images.
+    """
+    assert isinstance(name, str), name
+    assert isinstance(json_file, (str, os.PathLike)), json_file
+    assert isinstance(image_root, (str, os.PathLike)), image_root
+    # 1. register a function which returns dicts
+    DatasetCatalog.register(name, lambda: custom_load_coco_json(json_file, image_root, name))
+
+    # 2. Optionally, add metadata about this dataset,
+    # since they might be useful in evaluation, visualization or logging
+    MetadataCatalog.get(name).set(
+        json_file=json_file, image_root=image_root, evaluator_type="coco", **metadata
+    )
+
+
+def custom_register_coco_semseg(name, metadata, sem_seg_root, image_root):
+    assert isinstance(name, str), name
+    assert isinstance(sem_seg_root, (str, os.PathLike, list)), sem_seg_root
+    assert isinstance(image_root, (str, os.PathLike, list)), image_root
+    # 1. register a function which returns dicts
+    DatasetCatalog.register(name, lambda: custom_load_sem_seg_list(sem_seg_root, image_root))
+
+    # 2. Optionally, add metadata about this dataset,
+    # since they might be useful in evaluation, visualization or logging
+    MetadataCatalog.get(name).set(
+        sem_seg_root=sem_seg_root,
+        image_root=image_root,
+        evaluator_type="sem_seg",
+        ignore_label=255,
+        **metadata,
+    )

approach/ovod/APE/ape/data/datasets/d_cube.py

@@ -0,0 +1,275 @@
+import logging
+import os
+
+import pycocotools.mask as mask_util
+
+from detectron2.data import DatasetCatalog, MetadataCatalog
+from detectron2.data.datasets.builtin_meta import _get_coco_instances_meta
+from detectron2.data.datasets.lvis_v0_5_categories import LVIS_CATEGORIES as LVIS_V0_5_CATEGORIES
+from detectron2.data.datasets.lvis_v1_categories import LVIS_CATEGORIES as LVIS_V1_CATEGORIES
+from detectron2.structures import BoxMode
+from detectron2.utils.file_io import PathManager
+from fvcore.common.timer import Timer
+
+from .lvis_v1_coco_category_image_count import LVIS_V1_COCO_CATEGORY_IMAGE_COUNT
+
+"""
+This file contains functions to parse LVIS-format annotations into dicts in the
+"Detectron2 format".
+"""
+
+logger = logging.getLogger(__name__)
+
+__all__ = ["load_d3_json", "register_d3_instances"]
+
+
+def register_d3_instances(name, metadata, json_file, image_root, anno_root):
+    """
+    Register a dataset in LVIS's json annotation format for instance detection and segmentation.
+
+    Args:
+        name (str): a name that identifies the dataset, e.g. "lvis_v0.5_train".
+        metadata (dict): extra metadata associated with this dataset. It can be an empty dict.
+        json_file (str): path to the json instance annotation file.
+        image_root (str or path-like): directory which contains all the images.
+    """
+    DatasetCatalog.register(name, lambda: load_d3_json(json_file, image_root, anno_root, name))
+    MetadataCatalog.get(name).set(
+        json_file=json_file, image_root=image_root, evaluator_type="d3", **metadata
+    )
+
+
+def load_d3_json(json_file, image_root, anno_root, dataset_name=None, extra_annotation_keys=None):
+    """
+    Load a json file in LVIS's annotation format.
+
+    Args:
+        json_file (str): full path to the LVIS json annotation file.
+        image_root (str): the directory where the images in this json file exists.
+        dataset_name (str): the name of the dataset (e.g., "lvis_v0.5_train").
+            If provided, this function will put "thing_classes" into the metadata
+            associated with this dataset.
+        extra_annotation_keys (list[str]): list of per-annotation keys that should also be
+            loaded into the dataset dict (besides "bbox", "bbox_mode", "category_id",
+            "segmentation"). The values for these keys will be returned as-is.
+
+    Returns:
+        list[dict]: a list of dicts in Detectron2 standard format. (See
+        `Using Custom Datasets </tutorials/datasets.html>`_ )
+
+    Notes:
+        1. This function does not read the image files.
+           The results do not have the "image" field.
+    """
+    from d_cube import D3
+
+    timer = Timer()
+
+    d3 = D3(image_root, anno_root)
+
+    if timer.seconds() > 1:
+        logger.info("Loading d3 takes {:.2f} seconds.".format(timer.seconds()))
+
+    id_map = None
+    if dataset_name is not None:
+        meta = MetadataCatalog.get(dataset_name)
+        cat_ids = sorted(d3.get_sent_ids())
+        cats = d3.load_sents(cat_ids)
+        # The categories in a custom json file may not be sorted.
+        thing_classes = [c["raw_sent"] for c in sorted(cats, key=lambda x: x["id"])]
+        meta.thing_classes = thing_classes
+
+        # In COCO, certain category ids are artificially removed,
+        # and by convention they are always ignored.
+        # We deal with COCO's id issue and translate
+        # the category ids to contiguous ids in [0, 80).
+
+        # It works by looking at the "categories" field in the json, therefore
+        # if users' own json also have incontiguous ids, we'll
+        # apply this mapping as well but print a warning.
+        if not (min(cat_ids) == 1 and max(cat_ids) == len(cat_ids)):
+            if "coco" not in dataset_name:
+                logger.warning(
+                    """
+Category ids in annotations are not in [1, #categories]! We'll apply a mapping for you.
+"""
+                )
+        id_map = {v: i for i, v in enumerate(cat_ids)}
+        meta.thing_dataset_id_to_contiguous_id = id_map
+
+    img_ids = d3.get_img_ids()
+    imgs = d3.load_imgs(img_ids)
+    anno_ids = [d3.get_anno_ids(img_ids=img_id) for img_id in img_ids]
+    anns = [d3.load_annos(anno_ids=anno_id) for anno_id in anno_ids]
+    total_num_valid_anns = sum([len(x) for x in anns])
+    total_num_anns = len(d3.load_annos())
+    if total_num_valid_anns < total_num_anns:
+        logger.warning(
+            f"{anno_root} contains {total_num_anns} annotations, but only "
+            f"{total_num_valid_anns} of them match to images in the file."
+        )
+
+    imgs_anns = list(zip(imgs, anns))
+    logger.info("Loaded {} images in COCO format from {}".format(len(imgs_anns), json_file))
+
+    dataset_dicts = []
+
+    ann_keys = ["iscrowd", "bbox", "keypoints", "sent_id"] + (extra_annotation_keys or [])
+
+    num_instances_without_valid_segmentation = 0
+
+    for (img_dict, anno_dict_list) in imgs_anns:
+        record = {}
+        record["file_name"] = os.path.join(image_root, img_dict["file_name"])
+        record["height"] = img_dict["height"]
+        record["width"] = img_dict["width"]
+        image_id = record["image_id"] = img_dict["id"]
+
+        if meta.group == "intra":
+            group_ids = d3.get_group_ids(img_ids=[image_id])
+            sent_ids = d3.get_sent_ids(group_ids=group_ids)
+            sent_list = d3.load_sents(sent_ids=sent_ids)
+            # assert len(anno_dict_list) == len(sent_ids)
+        elif meta.group == "inter":
+            sent_ids = d3.get_sent_ids()
+            sent_list = d3.load_sents(sent_ids=sent_ids)
+            # sent_list = d3.load_sents()
+        else:
+            assert False
+        ref_list = [sent["raw_sent"] for sent in sent_list]
+        record["expressions"] = ref_list
+        if id_map:
+            record["sent_ids"] = [id_map[x] for x in sent_ids]
+
+        objs = []
+        for anno in anno_dict_list:
+            # Check that the image_id in this annotation is the same as
+            # the image_id we're looking at.
+            # This fails only when the data parsing logic or the annotation file is buggy.
+
+            # The original COCO valminusminival2014 & minival2014 annotation files
+            # actually contains bugs that, together with certain ways of using COCO API,
+            # can trigger this assertion.
+            assert anno["image_id"] == image_id
+
+            assert anno.get("ignore", 0) == 0, '"ignore" in COCO json file is not supported.'
+
+            obj = {key: anno[key] for key in ann_keys if key in anno}
+            if "bbox" in obj and len(obj["bbox"]) == 0:
+                raise ValueError(
+                    f"One annotation of image {image_id} contains empty 'bbox' value! "
+                    "This json does not have valid COCO format."
+                )
+
+            assert len(obj["bbox"]) == 1
+            obj["bbox"] = list(obj["bbox"][0])
+            # assert len(obj["sent_id"]) == 1
+            obj["sent_id"] = obj["sent_id"][0]
+
+            segm = anno.get("segmentation", None)
+            assert len(segm) == 1
+            segm = segm[0]
+            if segm:  # either list[list[float]] or dict(RLE)
+                if isinstance(segm, dict):
+                    if isinstance(segm["counts"], list):
+                        # convert to compressed RLE
+                        segm = mask_util.frPyObjects(segm, *segm["size"])
+                else:
+                    # filter out invalid polygons (< 3 points)
+                    segm = [poly for poly in segm if len(poly) % 2 == 0 and len(poly) >= 6]
+                    if len(segm) == 0:
+                        num_instances_without_valid_segmentation += 1
+                        continue  # ignore this instance
+                obj["segmentation"] = segm
+
+            keypts = anno.get("keypoints", None)
+            if keypts:  # list[int]
+                for idx, v in enumerate(keypts):
+                    if idx % 3 != 2:
+                        # COCO's segmentation coordinates are floating points in [0, H or W],
+                        # but keypoint coordinates are integers in [0, H-1 or W-1]
+                        # Therefore we assume the coordinates are "pixel indices" and
+                        # add 0.5 to convert to floating point coordinates.
+                        keypts[idx] = v + 0.5
+                obj["keypoints"] = keypts
+
+            obj["bbox_mode"] = BoxMode.XYWH_ABS
+            if id_map:
+                annotation_category_id = obj["sent_id"]
+                try:
+                    obj["sent_id"] = id_map[annotation_category_id]
+                except KeyError as e:
+                    raise KeyError(
+                        f"Encountered sent_id={annotation_category_id} "
+                        "but this id does not exist in 'categories' of the json file."
+                    ) from e
+            obj["category_id"] = obj["sent_id"]
+            obj["iscrowd"] = 0
+            objs.append(obj)
+        record["annotations"] = objs
+        dataset_dicts.append(record)
+
+    if num_instances_without_valid_segmentation > 0:
+        logger.warning(
+            "Filtered out {} instances without valid segmentation. ".format(
+                num_instances_without_valid_segmentation
+            )
+            + "There might be issues in your dataset generation process.  Please "
+            "check https://detectron2.readthedocs.io/en/latest/tutorials/datasets.html carefully"
+        )
+    return dataset_dicts
+
+
+def get_d3_instances_meta(dataset_name):
+    if "intra_scenario" in dataset_name:
+        group = "intra"
+    elif "inter_scenario" in dataset_name:
+        group = "inter"
+    else:
+        assert False
+    return {"group": group}
+
+
+_PREDEFINED_SPLITS_D3 = {
+    "d3_inter_scenario": {
+        "d3_inter_scenario": (
+            "D3/d3_images/",
+            {
+                "FULL": "D3/d3_json/d3_full_annotations.json",
+                "PRES": "D3/d3_json/d3_pres_annotations.json",
+                "ABS": "D3/d3_json/d3_abs_annotations.json",
+            },
+            "D3/d3_pkl/",
+        ),
+    },
+    "d3_intra_scenario": {
+        "d3_intra_scenario": (
+            "D3/d3_images/",
+            {
+                "FULL": "D3/d3_json/d3_full_annotations.json",
+                "PRES": "D3/d3_json/d3_pres_annotations.json",
+                "ABS": "D3/d3_json/d3_abs_annotations.json",
+            },
+            "D3/d3_pkl/",
+        ),
+    },
+}
+
+
+def register_all_D3(root):
+    for dataset_name, splits_per_dataset in _PREDEFINED_SPLITS_D3.items():
+        for key, (image_root, json_file, anno_root) in splits_per_dataset.items():
+            register_d3_instances(
+                key,
+                get_d3_instances_meta(dataset_name),
+                # os.path.join(root, json_file) if "://" not in json_file else json_file,
+                {k: os.path.join(root, v) for k, v in json_file.items()},
+                os.path.join(root, image_root),
+                os.path.join(root, anno_root),
+            )
+
+
+if __name__.endswith(".d_cube"):
+    # Assume pre-defined datasets live in `./datasets`.
+    _root = os.getenv("DETECTRON2_DATASETS", "datasets")
+    register_all_D3(_root)

approach/ovod/APE/ape/data/datasets/flickr30k.py

@@ -0,0 +1,68 @@
+import logging
+import os
+
+from .coco import custom_register_coco_instances
+
+logger = logging.getLogger(__name__)
+
+
+def _get_builtin_metadata(dataset_name):
+    return _get_flickr30k_metadata([])
+
+    raise KeyError("No built-in metadata for dataset {}".format(dataset_name))
+
+
+def _get_flickr30k_metadata(categories):
+    if len(categories) == 0:
+        return {}
+    id_to_name = {x["id"]: x["name"] for x in categories}
+    thing_dataset_id_to_contiguous_id = {i + 1: i for i in range(len(categories))}
+    thing_classes = [id_to_name[k] for k in sorted(id_to_name)]
+    return {
+        "thing_dataset_id_to_contiguous_id": thing_dataset_id_to_contiguous_id,
+        "thing_classes": thing_classes,
+    }
+
+
+_PREDEFINED_SPLITS_FLICKR30k = {}
+_PREDEFINED_SPLITS_FLICKR30k["flickr30k"] = {
+    "flickr30k": (
+        "flickr30k/flickr30k-images",
+        "flickr30k/flickr30k.json",
+    ),
+    "flickr30k_separateGT_train": (
+        "flickr30k/flickr30k-images",
+        "flickr30k/flickr30k_separateGT_train.json",
+    ),
+    "flickr30k_separateGT_val": (
+        "flickr30k/flickr30k-images",
+        "flickr30k/flickr30k_separateGT_val.json",
+    ),
+    "flickr30k_mergedGT_train": (
+        "flickr30k/flickr30k-images",
+        "flickr30k/flickr30k_mergedGT_train.json",
+    ),
+    "flickr30k_mergedGT_val": (
+        "flickr30k/flickr30k-images",
+        "flickr30k/flickr30k_mergedGT_val.json",
+    ),
+}
+
+
+def register_all_flickr30k(root):
+    for dataset_name, splits_per_dataset in _PREDEFINED_SPLITS_FLICKR30k.items():
+        for key, (image_root, json_file) in splits_per_dataset.items():
+            custom_register_coco_instances(
+                key,
+                _get_builtin_metadata(dataset_name),
+                os.path.join(root, json_file) if "://" not in json_file else json_file,
+                os.path.join(root, image_root),
+            )
+
+
+# True for open source;
+# Internally at fb, we register them elsewhere
+if __name__.endswith(".flickr30k"):
+    # Assume pre-defined datasets live in `./datasets`.
+    _root = os.path.expanduser(os.getenv("DETECTRON2_DATASETS", "datasets"))
+    register_all_flickr30k(_root)

approach/ovod/APE/ape/data/datasets/gqa.py

@@ -0,0 +1,60 @@
+import logging
+import os
+
+from .coco import custom_register_coco_instances
+
+logger = logging.getLogger(__name__)
+
+
+def _get_builtin_metadata(dataset_name):
+    return _get_gqa_metadata([])
+
+    raise KeyError("No built-in metadata for dataset {}".format(dataset_name))
+
+
+def _get_gqa_metadata(categories):
+    if len(categories) == 0:
+        return {}
+    id_to_name = {x["id"]: x["name"] for x in categories}
+    thing_dataset_id_to_contiguous_id = {i + 1: i for i in range(len(categories))}
+    thing_classes = [id_to_name[k] for k in sorted(id_to_name)]
+    return {
+        "thing_dataset_id_to_contiguous_id": thing_dataset_id_to_contiguous_id,
+        "thing_classes": thing_classes,
+    }
+
+
+_PREDEFINED_SPLITS_GQA = {}
+_PREDEFINED_SPLITS_GQA["gqa_region"] = {
+    "gqa_region": (
+        "gqa/images",
+        "gqa/gqa_region.json",
+    ),
+    "gqa_region_train": (
+        "gqa/images",
+        "gqa/gqa_region_train.json",
+    ),
+    "gqa_region_val": (
+        "gqa/images",
+        "gqa/gqa_region_val.json",
+    ),
+}
+
+
+def register_all_gqa(root):
+    for dataset_name, splits_per_dataset in _PREDEFINED_SPLITS_GQA.items():
+        for key, (image_root, json_file) in splits_per_dataset.items():
+            custom_register_coco_instances(
+                key,
+                _get_builtin_metadata(dataset_name),
+                os.path.join(root, json_file) if "://" not in json_file else json_file,
+                os.path.join(root, image_root),
+            )
+
+
+# True for open source;
+# Internally at fb, we register them elsewhere
+if __name__.endswith(".gqa"):
+    # Assume pre-defined datasets live in `./datasets`.
+    _root = os.path.expanduser(os.getenv("DETECTRON2_DATASETS", "datasets"))
+    register_all_gqa(_root)

approach/ovod/APE/ape/data/datasets/grit.py

@@ -0,0 +1,60 @@
+import os
+
+from .coco import custom_register_coco_instances
+
+GRIT_CATEGORIES = [
+    {"id": 0, "name": "object"},
+]
+
+
+def _get_builtin_metadata(dataset_name):
+    id_to_name = {x["id"]: x["name"] for x in GRIT_CATEGORIES}
+    thing_dataset_id_to_contiguous_id = {i: i for i in range(len(GRIT_CATEGORIES))}
+    thing_classes = [id_to_name[k] for k in sorted(id_to_name)]
+    return {
+        "thing_dataset_id_to_contiguous_id": thing_dataset_id_to_contiguous_id,
+        "thing_classes": thing_classes,
+    }
+
+
+_PREDEFINED_SPLITS_GRIT = {
+    "grit": ("GRIT/images", "GRIT/grit.json"),
+    "grit_0_snappy": ("GRIT/images", "GRIT/grit_0_snappy.json"),
+    "grit_1_snappy": ("GRIT/images", "GRIT/grit_1_snappy.json"),
+    "grit_2_snappy": ("GRIT/images", "GRIT/grit_2_snappy.json"),
+    "grit_3_snappy": ("GRIT/images", "GRIT/grit_3_snappy.json"),
+    "grit_4_snappy": ("GRIT/images", "GRIT/grit_4_snappy.json"),
+    "grit_5_snappy": ("GRIT/images", "GRIT/grit_5_snappy.json"),
+    "grit_6_snappy": ("GRIT/images", "GRIT/grit_6_snappy.json"),
+    "grit_7_snappy": ("GRIT/images", "GRIT/grit_7_snappy.json"),
+    "grit_8_snappy": ("GRIT/images", "GRIT/grit_8_snappy.json"),
+    "grit_9_snappy": ("GRIT/images", "GRIT/grit_9_snappy.json"),
+    "grit_10_snappy": ("GRIT/images", "GRIT/grit_10_snappy.json"),
+    "grit_11_snappy": ("GRIT/images", "GRIT/grit_11_snappy.json"),
+    "grit_12_snappy": ("GRIT/images", "GRIT/grit_12_snappy.json"),
+    "grit_13_snappy": ("GRIT/images", "GRIT/grit_13_snappy.json"),
+    "grit_14_snappy": ("GRIT/images", "GRIT/grit_14_snappy.json"),
+    "grit_15_snappy": ("GRIT/images", "GRIT/grit_15_snappy.json"),
+    "grit_16_snappy": ("GRIT/images", "GRIT/grit_16_snappy.json"),
+    "grit_17_snappy": ("GRIT/images", "GRIT/grit_17_snappy.json"),
+    "grit_18_snappy": ("GRIT/images", "GRIT/grit_18_snappy.json"),
+    "grit_19_snappy": ("GRIT/images", "GRIT/grit_19_snappy.json"),
+    "grit_20_snappy": ("GRIT/images", "GRIT/grit_20_snappy.json"),
+    "grit_21_snappy": ("GRIT/images", "GRIT/grit_21_snappy.json"),
+}
+
+
+def register_all_GRIT(root):
+    for key, (image_root, json_file) in _PREDEFINED_SPLITS_GRIT.items():
+        custom_register_coco_instances(
+            key,
+            _get_builtin_metadata(key),
+            os.path.join(root, json_file) if "://" not in json_file else json_file,
+            os.path.join(root, image_root),
+        )
+
+
+if __name__.endswith(".grit"):
+    # Assume pre-defined datasets live in `./datasets`.
+    _root = os.getenv("DETECTRON2_DATASETS", "datasets")
+    register_all_GRIT(_root)

approach/ovod/APE/ape/data/datasets/inst_categories.py

@@ -0,0 +1,488 @@
+categories = {
+    "coco": [
+        {"color": [220, 20, 60], "isthing": 1, "id": 1, "name": "person"},
+        {"color": [119, 11, 32], "isthing": 1, "id": 2, "name": "bicycle"},
+        {"color": [0, 0, 142], "isthing": 1, "id": 3, "name": "car"},
+        {"color": [0, 0, 230], "isthing": 1, "id": 4, "name": "motorcycle"},
+        {"color": [106, 0, 228], "isthing": 1, "id": 5, "name": "airplane"},
+        {"color": [0, 60, 100], "isthing": 1, "id": 6, "name": "bus"},
+        {"color": [0, 80, 100], "isthing": 1, "id": 7, "name": "train"},
+        {"color": [0, 0, 70], "isthing": 1, "id": 8, "name": "truck"},
+        {"color": [0, 0, 192], "isthing": 1, "id": 9, "name": "boat"},
+        {"color": [250, 170, 30], "isthing": 1, "id": 10, "name": "traffic light"},
+        {"color": [100, 170, 30], "isthing": 1, "id": 11, "name": "fire hydrant"},
+        {"color": [220, 220, 0], "isthing": 1, "id": 13, "name": "stop sign"},
+        {"color": [175, 116, 175], "isthing": 1, "id": 14, "name": "parking meter"},
+        {"color": [250, 0, 30], "isthing": 1, "id": 15, "name": "bench"},
+        {"color": [165, 42, 42], "isthing": 1, "id": 16, "name": "bird"},
+        {"color": [255, 77, 255], "isthing": 1, "id": 17, "name": "cat"},
+        {"color": [0, 226, 252], "isthing": 1, "id": 18, "name": "dog"},
+        {"color": [182, 182, 255], "isthing": 1, "id": 19, "name": "horse"},
+        {"color": [0, 82, 0], "isthing": 1, "id": 20, "name": "sheep"},
+        {"color": [120, 166, 157], "isthing": 1, "id": 21, "name": "cow"},
+        {"color": [110, 76, 0], "isthing": 1, "id": 22, "name": "elephant"},
+        {"color": [174, 57, 255], "isthing": 1, "id": 23, "name": "bear"},
+        {"color": [199, 100, 0], "isthing": 1, "id": 24, "name": "zebra"},
+        {"color": [72, 0, 118], "isthing": 1, "id": 25, "name": "giraffe"},
+        {"color": [255, 179, 240], "isthing": 1, "id": 27, "name": "backpack"},
+        {"color": [0, 125, 92], "isthing": 1, "id": 28, "name": "umbrella"},
+        {"color": [209, 0, 151], "isthing": 1, "id": 31, "name": "handbag"},
+        {"color": [188, 208, 182], "isthing": 1, "id": 32, "name": "tie"},
+        {"color": [0, 220, 176], "isthing": 1, "id": 33, "name": "suitcase"},
+        {"color": [255, 99, 164], "isthing": 1, "id": 34, "name": "frisbee"},
+        {"color": [92, 0, 73], "isthing": 1, "id": 35, "name": "skis"},
+        {"color": [133, 129, 255], "isthing": 1, "id": 36, "name": "snowboard"},
+        {"color": [78, 180, 255], "isthing": 1, "id": 37, "name": "sports ball"},
+        {"color": [0, 228, 0], "isthing": 1, "id": 38, "name": "kite"},
+        {"color": [174, 255, 243], "isthing": 1, "id": 39, "name": "baseball bat"},
+        {"color": [45, 89, 255], "isthing": 1, "id": 40, "name": "baseball glove"},
+        {"color": [134, 134, 103], "isthing": 1, "id": 41, "name": "skateboard"},
+        {"color": [145, 148, 174], "isthing": 1, "id": 42, "name": "surfboard"},
+        {"color": [255, 208, 186], "isthing": 1, "id": 43, "name": "tennis racket"},
+        {"color": [197, 226, 255], "isthing": 1, "id": 44, "name": "bottle"},
+        {"color": [171, 134, 1], "isthing": 1, "id": 46, "name": "wine glass"},
+        {"color": [109, 63, 54], "isthing": 1, "id": 47, "name": "cup"},
+        {"color": [207, 138, 255], "isthing": 1, "id": 48, "name": "fork"},
+        {"color": [151, 0, 95], "isthing": 1, "id": 49, "name": "knife"},
+        {"color": [9, 80, 61], "isthing": 1, "id": 50, "name": "spoon"},
+        {"color": [84, 105, 51], "isthing": 1, "id": 51, "name": "bowl"},
+        {"color": [74, 65, 105], "isthing": 1, "id": 52, "name": "banana"},
+        {"color": [166, 196, 102], "isthing": 1, "id": 53, "name": "apple"},
+        {"color": [208, 195, 210], "isthing": 1, "id": 54, "name": "sandwich"},
+        {"color": [255, 109, 65], "isthing": 1, "id": 55, "name": "orange"},
+        {"color": [0, 143, 149], "isthing": 1, "id": 56, "name": "broccoli"},
+        {"color": [179, 0, 194], "isthing": 1, "id": 57, "name": "carrot"},
+        {"color": [209, 99, 106], "isthing": 1, "id": 58, "name": "hot dog"},
+        {"color": [5, 121, 0], "isthing": 1, "id": 59, "name": "pizza"},
+        {"color": [227, 255, 205], "isthing": 1, "id": 60, "name": "donut"},
+        {"color": [147, 186, 208], "isthing": 1, "id": 61, "name": "cake"},
+        {"color": [153, 69, 1], "isthing": 1, "id": 62, "name": "chair"},
+        {"color": [3, 95, 161], "isthing": 1, "id": 63, "name": "couch"},
+        {"color": [163, 255, 0], "isthing": 1, "id": 64, "name": "potted plant"},
+        {"color": [119, 0, 170], "isthing": 1, "id": 65, "name": "bed"},
+        {"color": [0, 182, 199], "isthing": 1, "id": 67, "name": "dining table"},
+        {"color": [0, 165, 120], "isthing": 1, "id": 70, "name": "toilet"},
+        {"color": [183, 130, 88], "isthing": 1, "id": 72, "name": "tv"},
+        {"color": [95, 32, 0], "isthing": 1, "id": 73, "name": "laptop"},
+        {"color": [130, 114, 135], "isthing": 1, "id": 74, "name": "mouse"},
+        {"color": [110, 129, 133], "isthing": 1, "id": 75, "name": "remote"},
+        {"color": [166, 74, 118], "isthing": 1, "id": 76, "name": "keyboard"},
+        {"color": [219, 142, 185], "isthing": 1, "id": 77, "name": "cell phone"},
+        {"color": [79, 210, 114], "isthing": 1, "id": 78, "name": "microwave"},
+        {"color": [178, 90, 62], "isthing": 1, "id": 79, "name": "oven"},
+        {"color": [65, 70, 15], "isthing": 1, "id": 80, "name": "toaster"},
+        {"color": [127, 167, 115], "isthing": 1, "id": 81, "name": "sink"},
+        {"color": [59, 105, 106], "isthing": 1, "id": 82, "name": "refrigerator"},
+        {"color": [142, 108, 45], "isthing": 1, "id": 84, "name": "book"},
+        {"color": [196, 172, 0], "isthing": 1, "id": 85, "name": "clock"},
+        {"color": [95, 54, 80], "isthing": 1, "id": 86, "name": "vase"},
+        {"color": [128, 76, 255], "isthing": 1, "id": 87, "name": "scissors"},
+        {"color": [201, 57, 1], "isthing": 1, "id": 88, "name": "teddy bear"},
+        {"color": [246, 0, 122], "isthing": 1, "id": 89, "name": "hair drier"},
+        {"color": [191, 162, 208], "isthing": 1, "id": 90, "name": "toothbrush"},
+    ],
+    "cityscapes": [
+        {"id": i + 1, "name": x}
+        for i, x in enumerate(
+            ["person", "rider", "car", "truck", "bus", "train", "motorcycle", "bicycle"]
+        )
+    ],
+    "mapillary": [
+        {"id": 1, "name": "animal--bird"},
+        {"id": 2, "name": "animal--ground-animal"},
+        {"id": 9, "name": "construction--flat--crosswalk-plain"},
+        {"id": 20, "name": "human--person"},
+        {"id": 21, "name": "human--rider--bicyclist"},
+        {"id": 22, "name": "human--rider--motorcyclist"},
+        {"id": 23, "name": "human--rider--other-rider"},
+        {"id": 24, "name": "marking--crosswalk-zebra"},
+        {"id": 33, "name": "object--banner"},
+        {"id": 34, "name": "object--bench"},
+        {"id": 35, "name": "object--bike-rack"},
+        {"id": 36, "name": "object--billboard"},
+        {"id": 37, "name": "object--catch-basin"},
+        {"id": 38, "name": "object--cctv-camera"},
+        {"id": 39, "name": "object--fire-hydrant"},
+        {"id": 40, "name": "object--junction-box"},
+        {"id": 41, "name": "object--mailbox"},
+        {"id": 42, "name": "object--manhole"},
+        {"id": 43, "name": "object--phone-booth"},
+        {"id": 45, "name": "object--street-light"},
+        {"id": 46, "name": "object--support--pole"},
+        {"id": 47, "name": "object--support--traffic-sign-frame"},
+        {"id": 48, "name": "object--support--utility-pole"},
+        {"id": 49, "name": "object--traffic-light"},
+        {"id": 50, "name": "object--traffic-sign--back"},
+        {"id": 51, "name": "object--traffic-sign--front"},
+        {"id": 52, "name": "object--trash-can"},
+        {"id": 53, "name": "object--vehicle--bicycle"},
+        {"id": 54, "name": "object--vehicle--boat"},
+        {"id": 55, "name": "object--vehicle--bus"},
+        {"id": 56, "name": "object--vehicle--car"},
+        {"id": 57, "name": "object--vehicle--caravan"},
+        {"id": 58, "name": "object--vehicle--motorcycle"},
+        {"id": 60, "name": "object--vehicle--other-vehicle"},
+        {"id": 61, "name": "object--vehicle--trailer"},
+        {"id": 62, "name": "object--vehicle--truck"},
+        {"id": 63, "name": "object--vehicle--wheeled-slow"},
+    ],
+    "viper": [
+        {"id": 13, "name": "trafficlight", "supercategory": ""},
+        {"id": 16, "name": "firehydrant", "supercategory": ""},
+        {"id": 17, "name": "chair", "supercategory": ""},
+        {"id": 19, "name": "trashcan", "supercategory": ""},
+        {"id": 20, "name": "person", "supercategory": ""},
+        {"id": 23, "name": "motorcycle", "supercategory": ""},
+        {"id": 24, "name": "car", "supercategory": ""},
+        {"id": 25, "name": "van", "supercategory": ""},
+        {"id": 26, "name": "bus", "supercategory": ""},
+        {"id": 27, "name": "truck", "supercategory": ""},
+    ],
+    "scannet": [
+        {"id": 3, "name": "cabinet", "supercategory": "furniture"},
+        {"id": 4, "name": "bed", "supercategory": "furniture"},
+        {"id": 5, "name": "chair", "supercategory": "furniture"},
+        {"id": 6, "name": "sofa", "supercategory": "furniture"},
+        {"id": 7, "name": "table", "supercategory": "furniture"},
+        {"id": 8, "name": "door", "supercategory": "furniture"},
+        {"id": 9, "name": "window", "supercategory": "furniture"},
+        {"id": 10, "name": "bookshelf", "supercategory": "furniture"},
+        {"id": 11, "name": "picture", "supercategory": "furniture"},
+        {"id": 12, "name": "counter", "supercategory": "furniture"},
+        {"id": 14, "name": "desk", "supercategory": "furniture"},
+        {"id": 16, "name": "curtain", "supercategory": "furniture"},
+        {"id": 24, "name": "refrigerator", "supercategory": "appliance"},
+        {"id": 28, "name": "shower curtain", "supercategory": "furniture"},
+        {"id": 33, "name": "toilet", "supercategory": "furniture"},
+        {"id": 34, "name": "sink", "supercategory": "appliance"},
+        {"id": 36, "name": "bathtub", "supercategory": "furniture"},
+        {"id": 39, "name": "otherfurniture", "supercategory": "furniture"},
+    ],
+    "oid": [
+        {"id": 1, "name": "Screwdriver", "freebase_id": "/m/01bms0"},
+        {"id": 2, "name": "Light switch", "freebase_id": "/m/03jbxj"},
+        {"id": 3, "name": "Doughnut", "freebase_id": "/m/0jy4k"},
+        {"id": 4, "name": "Toilet paper", "freebase_id": "/m/09gtd"},
+        {"id": 5, "name": "Wrench", "freebase_id": "/m/01j5ks"},
+        {"id": 6, "name": "Toaster", "freebase_id": "/m/01k6s3"},
+        {"id": 7, "name": "Tennis ball", "freebase_id": "/m/05ctyq"},
+        {"id": 8, "name": "Radish", "freebase_id": "/m/015x5n"},
+        {"id": 9, "name": "Pomegranate", "freebase_id": "/m/0jwn_"},
+        {"id": 10, "name": "Kite", "freebase_id": "/m/02zt3"},
+        {"id": 11, "name": "Table tennis racket", "freebase_id": "/m/05_5p_0"},
+        {"id": 12, "name": "Hamster", "freebase_id": "/m/03qrc"},
+        {"id": 13, "name": "Barge", "freebase_id": "/m/01btn"},
+        {"id": 14, "name": "Shower", "freebase_id": "/m/02f9f_"},
+        {"id": 15, "name": "Printer", "freebase_id": "/m/01m4t"},
+        {"id": 16, "name": "Snowmobile", "freebase_id": "/m/01x3jk"},
+        {"id": 17, "name": "Fire hydrant", "freebase_id": "/m/01pns0"},
+        {"id": 18, "name": "Limousine", "freebase_id": "/m/01lcw4"},
+        {"id": 19, "name": "Whale", "freebase_id": "/m/084zz"},
+        {"id": 20, "name": "Microwave oven", "freebase_id": "/m/0fx9l"},
+        {"id": 21, "name": "Asparagus", "freebase_id": "/m/0cjs7"},
+        {"id": 22, "name": "Lion", "freebase_id": "/m/096mb"},
+        {"id": 23, "name": "Spatula", "freebase_id": "/m/02d1br"},
+        {"id": 24, "name": "Torch", "freebase_id": "/m/07dd4"},
+        {"id": 25, "name": "Volleyball", "freebase_id": "/m/02rgn06"},
+        {"id": 26, "name": "Ambulance", "freebase_id": "/m/012n7d"},
+        {"id": 27, "name": "Chopsticks", "freebase_id": "/m/01_5g"},
+        {"id": 28, "name": "Raccoon", "freebase_id": "/m/0dq75"},
+        {"id": 29, "name": "Blue jay", "freebase_id": "/m/01f8m5"},
+        {"id": 30, "name": "Lynx", "freebase_id": "/m/04g2r"},
+        {"id": 31, "name": "Dice", "freebase_id": "/m/029b3"},
+        {"id": 32, "name": "Filing cabinet", "freebase_id": "/m/047j0r"},
+        {"id": 33, "name": "Ruler", "freebase_id": "/m/0hdln"},
+        {"id": 34, "name": "Power plugs and sockets", "freebase_id": "/m/03bbps"},
+        {"id": 35, "name": "Bell pepper", "freebase_id": "/m/0jg57"},
+        {"id": 36, "name": "Binoculars", "freebase_id": "/m/0lt4_"},
+        {"id": 37, "name": "Pretzel", "freebase_id": "/m/01f91_"},
+        {"id": 38, "name": "Hot dog", "freebase_id": "/m/01b9xk"},
+        {"id": 39, "name": "Missile", "freebase_id": "/m/04ylt"},
+        {"id": 40, "name": "Common fig", "freebase_id": "/m/043nyj"},
+        {"id": 41, "name": "Croissant", "freebase_id": "/m/015wgc"},
+        {"id": 42, "name": "Adhesive tape", "freebase_id": "/m/03m3vtv"},
+        {"id": 43, "name": "Slow cooker", "freebase_id": "/m/02tsc9"},
+        {"id": 44, "name": "Dog bed", "freebase_id": "/m/0h8n6f9"},
+        {"id": 45, "name": "Harpsichord", "freebase_id": "/m/03q5t"},
+        {"id": 46, "name": "Billiard table", "freebase_id": "/m/04p0qw"},
+        {"id": 47, "name": "Alpaca", "freebase_id": "/m/0pcr"},
+        {"id": 48, "name": "Harbor seal", "freebase_id": "/m/02l8p9"},
+        {"id": 49, "name": "Grape", "freebase_id": "/m/0388q"},
+        {"id": 50, "name": "Nail", "freebase_id": "/m/05bm6"},
+        {"id": 51, "name": "Paper towel", "freebase_id": "/m/02w3r3"},
+        {"id": 52, "name": "Alarm clock", "freebase_id": "/m/046dlr"},
+        {"id": 53, "name": "Guacamole", "freebase_id": "/m/02g30s"},
+        {"id": 54, "name": "Starfish", "freebase_id": "/m/01h8tj"},
+        {"id": 55, "name": "Zebra", "freebase_id": "/m/0898b"},
+        {"id": 56, "name": "Segway", "freebase_id": "/m/076bq"},
+        {"id": 57, "name": "Sea turtle", "freebase_id": "/m/0120dh"},
+        {"id": 58, "name": "Scissors", "freebase_id": "/m/01lsmm"},
+        {"id": 59, "name": "Rhinoceros", "freebase_id": "/m/03d443"},
+        {"id": 60, "name": "Kangaroo", "freebase_id": "/m/04c0y"},
+        {"id": 61, "name": "Jaguar", "freebase_id": "/m/0449p"},
+        {"id": 62, "name": "Leopard", "freebase_id": "/m/0c29q"},
+        {"id": 63, "name": "Dumbbell", "freebase_id": "/m/04h8sr"},
+        {"id": 64, "name": "Envelope", "freebase_id": "/m/0frqm"},
+        {"id": 65, "name": "Winter melon", "freebase_id": "/m/02cvgx"},
+        {"id": 66, "name": "Teapot", "freebase_id": "/m/01fh4r"},
+        {"id": 67, "name": "Camel", "freebase_id": "/m/01x_v"},
+        {"id": 68, "name": "Beaker", "freebase_id": "/m/0d20w4"},
+        {"id": 69, "name": "Brown bear", "freebase_id": "/m/01dxs"},
+        {"id": 70, "name": "Toilet", "freebase_id": "/m/09g1w"},
+        {"id": 71, "name": "Teddy bear", "freebase_id": "/m/0kmg4"},
+        {"id": 72, "name": "Briefcase", "freebase_id": "/m/0584n8"},
+        {"id": 73, "name": "Stop sign", "freebase_id": "/m/02pv19"},
+        {"id": 74, "name": "Tiger", "freebase_id": "/m/07dm6"},
+        {"id": 75, "name": "Cabbage", "freebase_id": "/m/0fbw6"},
+        {"id": 76, "name": "Giraffe", "freebase_id": "/m/03bk1"},
+        {"id": 77, "name": "Polar bear", "freebase_id": "/m/0633h"},
+        {"id": 78, "name": "Shark", "freebase_id": "/m/0by6g"},
+        {"id": 79, "name": "Rabbit", "freebase_id": "/m/06mf6"},
+        {"id": 80, "name": "Swim cap", "freebase_id": "/m/04tn4x"},
+        {"id": 81, "name": "Pressure cooker", "freebase_id": "/m/0h8ntjv"},
+        {"id": 82, "name": "Kitchen knife", "freebase_id": "/m/058qzx"},
+        {"id": 83, "name": "Submarine sandwich", "freebase_id": "/m/06pcq"},
+        {"id": 84, "name": "Flashlight", "freebase_id": "/m/01kb5b"},
+        {"id": 85, "name": "Penguin", "freebase_id": "/m/05z6w"},
+        {"id": 86, "name": "Snake", "freebase_id": "/m/078jl"},
+        {"id": 87, "name": "Zucchini", "freebase_id": "/m/027pcv"},
+        {"id": 88, "name": "Bat", "freebase_id": "/m/01h44"},
+        {"id": 89, "name": "Food processor", "freebase_id": "/m/03y6mg"},
+        {"id": 90, "name": "Ostrich", "freebase_id": "/m/05n4y"},
+        {"id": 91, "name": "Sea lion", "freebase_id": "/m/0gd36"},
+        {"id": 92, "name": "Goldfish", "freebase_id": "/m/03fj2"},
+        {"id": 93, "name": "Elephant", "freebase_id": "/m/0bwd_0j"},
+        {"id": 94, "name": "Rocket", "freebase_id": "/m/09rvcxw"},
+        {"id": 95, "name": "Mouse", "freebase_id": "/m/04rmv"},
+        {"id": 96, "name": "Oyster", "freebase_id": "/m/0_cp5"},
+        {"id": 97, "name": "Digital clock", "freebase_id": "/m/06_72j"},
+        {"id": 98, "name": "Otter", "freebase_id": "/m/0cn6p"},
+        {"id": 99, "name": "Dolphin", "freebase_id": "/m/02hj4"},
+        {"id": 100, "name": "Punching bag", "freebase_id": "/m/0420v5"},
+        {"id": 101, "name": "Corded phone", "freebase_id": "/m/0h8lkj8"},
+        {"id": 102, "name": "Tennis racket", "freebase_id": "/m/0h8my_4"},
+        {"id": 103, "name": "Pancake", "freebase_id": "/m/01dwwc"},
+        {"id": 104, "name": "Mango", "freebase_id": "/m/0fldg"},
+        {"id": 105, "name": "Crocodile", "freebase_id": "/m/09f_2"},
+        {"id": 106, "name": "Waffle", "freebase_id": "/m/01dwsz"},
+        {"id": 107, "name": "Computer mouse", "freebase_id": "/m/020lf"},
+        {"id": 108, "name": "Kettle", "freebase_id": "/m/03s_tn"},
+        {"id": 109, "name": "Tart", "freebase_id": "/m/02zvsm"},
+        {"id": 110, "name": "Oven", "freebase_id": "/m/029bxz"},
+        {"id": 111, "name": "Banana", "freebase_id": "/m/09qck"},
+        {"id": 112, "name": "Cheetah", "freebase_id": "/m/0cd4d"},
+        {"id": 113, "name": "Raven", "freebase_id": "/m/06j2d"},
+        {"id": 114, "name": "Frying pan", "freebase_id": "/m/04v6l4"},
+        {"id": 115, "name": "Pear", "freebase_id": "/m/061_f"},
+        {"id": 116, "name": "Fox", "freebase_id": "/m/0306r"},
+        {"id": 117, "name": "Skateboard", "freebase_id": "/m/06_fw"},
+        {"id": 118, "name": "Rugby ball", "freebase_id": "/m/0wdt60w"},
+        {"id": 119, "name": "Watermelon", "freebase_id": "/m/0kpqd"},
+        {"id": 120, "name": "Flute", "freebase_id": "/m/0l14j_"},
+        {"id": 121, "name": "Canary", "freebase_id": "/m/0ccs93"},
+        {"id": 122, "name": "Door handle", "freebase_id": "/m/03c7gz"},
+        {"id": 123, "name": "Saxophone", "freebase_id": "/m/06ncr"},
+        {"id": 124, "name": "Burrito", "freebase_id": "/m/01j3zr"},
+        {"id": 125, "name": "Suitcase", "freebase_id": "/m/01s55n"},
+        {"id": 126, "name": "Roller skates", "freebase_id": "/m/02p3w7d"},
+        {"id": 127, "name": "Dagger", "freebase_id": "/m/02gzp"},
+        {"id": 128, "name": "Seat belt", "freebase_id": "/m/0dkzw"},
+        {"id": 129, "name": "Washing machine", "freebase_id": "/m/0174k2"},
+        {"id": 130, "name": "Jet ski", "freebase_id": "/m/01xs3r"},
+        {"id": 131, "name": "Sombrero", "freebase_id": "/m/02jfl0"},
+        {"id": 132, "name": "Pig", "freebase_id": "/m/068zj"},
+        {"id": 133, "name": "Drinking straw", "freebase_id": "/m/03v5tg"},
+        {"id": 134, "name": "Peach", "freebase_id": "/m/0dj6p"},
+        {"id": 135, "name": "Tortoise", "freebase_id": "/m/011k07"},
+        {"id": 136, "name": "Towel", "freebase_id": "/m/0162_1"},
+        {"id": 137, "name": "Tablet computer", "freebase_id": "/m/0bh9flk"},
+        {"id": 138, "name": "Cucumber", "freebase_id": "/m/015x4r"},
+        {"id": 139, "name": "Mule", "freebase_id": "/m/0dbzx"},
+        {"id": 140, "name": "Potato", "freebase_id": "/m/05vtc"},
+        {"id": 141, "name": "Frog", "freebase_id": "/m/09ld4"},
+        {"id": 142, "name": "Bear", "freebase_id": "/m/01dws"},
+        {"id": 143, "name": "Lighthouse", "freebase_id": "/m/04h7h"},
+        {"id": 144, "name": "Belt", "freebase_id": "/m/0176mf"},
+        {"id": 145, "name": "Baseball bat", "freebase_id": "/m/03g8mr"},
+        {"id": 146, "name": "Racket", "freebase_id": "/m/0dv9c"},
+        {"id": 147, "name": "Sword", "freebase_id": "/m/06y5r"},
+        {"id": 148, "name": "Bagel", "freebase_id": "/m/01fb_0"},
+        {"id": 149, "name": "Goat", "freebase_id": "/m/03fwl"},
+        {"id": 150, "name": "Lizard", "freebase_id": "/m/04m9y"},
+        {"id": 151, "name": "Parrot", "freebase_id": "/m/0gv1x"},
+        {"id": 152, "name": "Owl", "freebase_id": "/m/09d5_"},
+        {"id": 153, "name": "Turkey", "freebase_id": "/m/0jly1"},
+        {"id": 154, "name": "Cello", "freebase_id": "/m/01xqw"},
+        {"id": 155, "name": "Knife", "freebase_id": "/m/04ctx"},
+        {"id": 156, "name": "Handgun", "freebase_id": "/m/0gxl3"},
+        {"id": 157, "name": "Carrot", "freebase_id": "/m/0fj52s"},
+        {"id": 158, "name": "Hamburger", "freebase_id": "/m/0cdn1"},
+        {"id": 159, "name": "Grapefruit", "freebase_id": "/m/0hqkz"},
+        {"id": 160, "name": "Tap", "freebase_id": "/m/02jz0l"},
+        {"id": 161, "name": "Tea", "freebase_id": "/m/07clx"},
+        {"id": 162, "name": "Bull", "freebase_id": "/m/0cnyhnx"},
+        {"id": 163, "name": "Turtle", "freebase_id": "/m/09dzg"},
+        {"id": 164, "name": "Bust", "freebase_id": "/m/04yqq2"},
+        {"id": 165, "name": "Monkey", "freebase_id": "/m/08pbxl"},
+        {"id": 166, "name": "Wok", "freebase_id": "/m/084rd"},
+        {"id": 167, "name": "Broccoli", "freebase_id": "/m/0hkxq"},
+        {"id": 168, "name": "Pitcher", "freebase_id": "/m/054fyh"},
+        {"id": 169, "name": "Whiteboard", "freebase_id": "/m/02d9qx"},
+        {"id": 170, "name": "Squirrel", "freebase_id": "/m/071qp"},
+        {"id": 171, "name": "Jug", "freebase_id": "/m/08hvt4"},
+        {"id": 172, "name": "Woodpecker", "freebase_id": "/m/01dy8n"},
+        {"id": 173, "name": "Pizza", "freebase_id": "/m/0663v"},
+        {"id": 174, "name": "Surfboard", "freebase_id": "/m/019w40"},
+        {"id": 175, "name": "Sofa bed", "freebase_id": "/m/03m3pdh"},
+        {"id": 176, "name": "Sheep", "freebase_id": "/m/07bgp"},
+        {"id": 177, "name": "Candle", "freebase_id": "/m/0c06p"},
+        {"id": 178, "name": "Muffin", "freebase_id": "/m/01tcjp"},
+        {"id": 179, "name": "Cookie", "freebase_id": "/m/021mn"},
+        {"id": 180, "name": "Apple", "freebase_id": "/m/014j1m"},
+        {"id": 181, "name": "Chest of drawers", "freebase_id": "/m/05kyg_"},
+        {"id": 182, "name": "Skull", "freebase_id": "/m/016m2d"},
+        {"id": 183, "name": "Chicken", "freebase_id": "/m/09b5t"},
+        {"id": 184, "name": "Loveseat", "freebase_id": "/m/0703r8"},
+        {"id": 185, "name": "Baseball glove", "freebase_id": "/m/03grzl"},
+        {"id": 186, "name": "Piano", "freebase_id": "/m/05r5c"},
+        {"id": 187, "name": "Waste container", "freebase_id": "/m/0bjyj5"},
+        {"id": 188, "name": "Barrel", "freebase_id": "/m/02zn6n"},
+        {"id": 189, "name": "Swan", "freebase_id": "/m/0dftk"},
+        {"id": 190, "name": "Taxi", "freebase_id": "/m/0pg52"},
+        {"id": 191, "name": "Lemon", "freebase_id": "/m/09k_b"},
+        {"id": 192, "name": "Pumpkin", "freebase_id": "/m/05zsy"},
+        {"id": 193, "name": "Sparrow", "freebase_id": "/m/0h23m"},
+        {"id": 194, "name": "Orange", "freebase_id": "/m/0cyhj_"},
+        {"id": 195, "name": "Tank", "freebase_id": "/m/07cmd"},
+        {"id": 196, "name": "Sandwich", "freebase_id": "/m/0l515"},
+        {"id": 197, "name": "Coffee", "freebase_id": "/m/02vqfm"},
+        {"id": 198, "name": "Juice", "freebase_id": "/m/01z1kdw"},
+        {"id": 199, "name": "Coin", "freebase_id": "/m/0242l"},
+        {"id": 200, "name": "Pen", "freebase_id": "/m/0k1tl"},
+        {"id": 201, "name": "Watch", "freebase_id": "/m/0gjkl"},
+        {"id": 202, "name": "Eagle", "freebase_id": "/m/09csl"},
+        {"id": 203, "name": "Goose", "freebase_id": "/m/0dbvp"},
+        {"id": 204, "name": "Falcon", "freebase_id": "/m/0f6wt"},
+        {"id": 205, "name": "Christmas tree", "freebase_id": "/m/025nd"},
+        {"id": 206, "name": "Sunflower", "freebase_id": "/m/0ftb8"},
+        {"id": 207, "name": "Vase", "freebase_id": "/m/02s195"},
+        {"id": 208, "name": "Football", "freebase_id": "/m/01226z"},
+        {"id": 209, "name": "Canoe", "freebase_id": "/m/0ph39"},
+        {"id": 210, "name": "High heels", "freebase_id": "/m/06k2mb"},
+        {"id": 211, "name": "Spoon", "freebase_id": "/m/0cmx8"},
+        {"id": 212, "name": "Mug", "freebase_id": "/m/02jvh9"},
+        {"id": 213, "name": "Swimwear", "freebase_id": "/m/01gkx_"},
+        {"id": 214, "name": "Duck", "freebase_id": "/m/09ddx"},
+        {"id": 215, "name": "Cat", "freebase_id": "/m/01yrx"},
+        {"id": 216, "name": "Tomato", "freebase_id": "/m/07j87"},
+        {"id": 217, "name": "Cocktail", "freebase_id": "/m/024g6"},
+        {"id": 218, "name": "Clock", "freebase_id": "/m/01x3z"},
+        {"id": 219, "name": "Cowboy hat", "freebase_id": "/m/025rp__"},
+        {"id": 220, "name": "Miniskirt", "freebase_id": "/m/01cmb2"},
+        {"id": 221, "name": "Cattle", "freebase_id": "/m/01xq0k1"},
+        {"id": 222, "name": "Strawberry", "freebase_id": "/m/07fbm7"},
+        {"id": 223, "name": "Bronze sculpture", "freebase_id": "/m/01yx86"},
+        {"id": 224, "name": "Pillow", "freebase_id": "/m/034c16"},
+        {"id": 225, "name": "Squash", "freebase_id": "/m/0dv77"},
+        {"id": 226, "name": "Traffic light", "freebase_id": "/m/015qff"},
+        {"id": 227, "name": "Saucer", "freebase_id": "/m/03q5c7"},
+        {"id": 228, "name": "Reptile", "freebase_id": "/m/06bt6"},
+        {"id": 229, "name": "Cake", "freebase_id": "/m/0fszt"},
+        {"id": 230, "name": "Plastic bag", "freebase_id": "/m/05gqfk"},
+        {"id": 231, "name": "Studio couch", "freebase_id": "/m/026qbn5"},
+        {"id": 232, "name": "Beer", "freebase_id": "/m/01599"},
+        {"id": 233, "name": "Scarf", "freebase_id": "/m/02h19r"},
+        {"id": 234, "name": "Coffee cup", "freebase_id": "/m/02p5f1q"},
+        {"id": 235, "name": "Wine", "freebase_id": "/m/081qc"},
+        {"id": 236, "name": "Mushroom", "freebase_id": "/m/052sf"},
+        {"id": 237, "name": "Traffic sign", "freebase_id": "/m/01mqdt"},
+        {"id": 238, "name": "Camera", "freebase_id": "/m/0dv5r"},
+        {"id": 239, "name": "Rose", "freebase_id": "/m/06m11"},
+        {"id": 240, "name": "Couch", "freebase_id": "/m/02crq1"},
+        {"id": 241, "name": "Handbag", "freebase_id": "/m/080hkjn"},
+        {"id": 242, "name": "Fedora", "freebase_id": "/m/02fq_6"},
+        {"id": 243, "name": "Sock", "freebase_id": "/m/01nq26"},
+        {"id": 244, "name": "Computer keyboard", "freebase_id": "/m/01m2v"},
+        {"id": 245, "name": "Mobile phone", "freebase_id": "/m/050k8"},
+        {"id": 246, "name": "Ball", "freebase_id": "/m/018xm"},
+        {"id": 247, "name": "Balloon", "freebase_id": "/m/01j51"},
+        {"id": 248, "name": "Horse", "freebase_id": "/m/03k3r"},
+        {"id": 249, "name": "Boot", "freebase_id": "/m/01b638"},
+        {"id": 250, "name": "Fish", "freebase_id": "/m/0ch_cf"},
+        {"id": 251, "name": "Backpack", "freebase_id": "/m/01940j"},
+        {"id": 252, "name": "Skirt", "freebase_id": "/m/02wv6h6"},
+        {"id": 253, "name": "Van", "freebase_id": "/m/0h2r6"},
+        {"id": 254, "name": "Bread", "freebase_id": "/m/09728"},
+        {"id": 255, "name": "Glove", "freebase_id": "/m/0174n1"},
+        {"id": 256, "name": "Dog", "freebase_id": "/m/0bt9lr"},
+        {"id": 257, "name": "Airplane", "freebase_id": "/m/0cmf2"},
+        {"id": 258, "name": "Motorcycle", "freebase_id": "/m/04_sv"},
+        {"id": 259, "name": "Drink", "freebase_id": "/m/0271t"},
+        {"id": 260, "name": "Book", "freebase_id": "/m/0bt_c3"},
+        {"id": 261, "name": "Train", "freebase_id": "/m/07jdr"},
+        {"id": 262, "name": "Flower", "freebase_id": "/m/0c9ph5"},
+        {"id": 263, "name": "Carnivore", "freebase_id": "/m/01lrl"},
+        {"id": 264, "name": "Human ear", "freebase_id": "/m/039xj_"},
+        {"id": 265, "name": "Toy", "freebase_id": "/m/0138tl"},
+        {"id": 266, "name": "Box", "freebase_id": "/m/025dyy"},
+        {"id": 267, "name": "Truck", "freebase_id": "/m/07r04"},
+        {"id": 268, "name": "Wheel", "freebase_id": "/m/083wq"},
+        {"id": 269, "name": "Aircraft", "freebase_id": "/m/0k5j"},
+        {"id": 270, "name": "Bus", "freebase_id": "/m/01bjv"},
+        {"id": 271, "name": "Human mouth", "freebase_id": "/m/0283dt1"},
+        {"id": 272, "name": "Sculpture", "freebase_id": "/m/06msq"},
+        {"id": 273, "name": "Shirt", "freebase_id": "/m/01n4qj"},
+        {"id": 274, "name": "Hat", "freebase_id": "/m/02dl1y"},
+        {"id": 275, "name": "Vehicle registration plate", "freebase_id": "/m/01jfm_"},
+        {"id": 276, "name": "Guitar", "freebase_id": "/m/0342h"},
+        {"id": 277, "name": "Sun hat", "freebase_id": "/m/02wbtzl"},
+        {"id": 278, "name": "Bottle", "freebase_id": "/m/04dr76w"},
+        {"id": 279, "name": "Luggage and bags", "freebase_id": "/m/0hf58v5"},
+        {"id": 280, "name": "Trousers", "freebase_id": "/m/07mhn"},
+        {"id": 281, "name": "Bicycle wheel", "freebase_id": "/m/01bqk0"},
+        {"id": 282, "name": "Suit", "freebase_id": "/m/01xyhv"},
+        {"id": 283, "name": "Bowl", "freebase_id": "/m/04kkgm"},
+        {"id": 284, "name": "Man", "freebase_id": "/m/04yx4"},
+        {"id": 285, "name": "Flowerpot", "freebase_id": "/m/0fm3zh"},
+        {"id": 286, "name": "Laptop", "freebase_id": "/m/01c648"},
+        {"id": 287, "name": "Boy", "freebase_id": "/m/01bl7v"},
+        {"id": 288, "name": "Picture frame", "freebase_id": "/m/06z37_"},
+        {"id": 289, "name": "Bird", "freebase_id": "/m/015p6"},
+        {"id": 290, "name": "Car", "freebase_id": "/m/0k4j"},
+        {"id": 291, "name": "Shorts", "freebase_id": "/m/01bfm9"},
+        {"id": 292, "name": "Woman", "freebase_id": "/m/03bt1vf"},
+        {"id": 293, "name": "Platter", "freebase_id": "/m/099ssp"},
+        {"id": 294, "name": "Tie", "freebase_id": "/m/01rkbr"},
+        {"id": 295, "name": "Girl", "freebase_id": "/m/05r655"},
+        {"id": 296, "name": "Skyscraper", "freebase_id": "/m/079cl"},
+        {"id": 297, "name": "Person", "freebase_id": "/m/01g317"},
+        {"id": 298, "name": "Flag", "freebase_id": "/m/03120"},
+        {"id": 299, "name": "Jeans", "freebase_id": "/m/0fly7"},
+        {"id": 300, "name": "Dress", "freebase_id": "/m/01d40f"},
+    ],
+    "kitti": [
+        {"id": 24, "name": "person"},
+        {"id": 25, "name": "rider"},
+        {"id": 26, "name": "car"},
+        {"id": 27, "name": "truck"},
+        {"id": 28, "name": "bus"},
+        {"id": 31, "name": "train"},
+        {"id": 32, "name": "motorcycle"},
+        {"id": 33, "name": "bicycle"},
+    ],
+    "wilddash": [
+        {"id": 1, "name": "ego vehicle"},
+        {"id": 24, "name": "person"},
+        {"id": 25, "name": "rider"},
+        {"id": 26, "name": "car"},
+        {"id": 27, "name": "truck"},
+        {"id": 28, "name": "bus"},
+        {"id": 29, "name": "caravan"},
+        {"id": 30, "name": "trailer"},
+        {"id": 31, "name": "train"},
+        {"id": 32, "name": "motorcycle"},
+        {"id": 33, "name": "bicycle"},
+        {"id": 34, "name": "pickup"},
+        {"id": 35, "name": "van"},
+    ],
+}

approach/ovod/APE/ape/data/datasets/lvis_coco.py

@@ -0,0 +1,284 @@
+import logging
+import os
+
+import pycocotools.mask as mask_util
+
+from detectron2.data import DatasetCatalog, MetadataCatalog
+from detectron2.data.datasets.builtin_meta import _get_coco_instances_meta
+from detectron2.data.datasets.lvis_v0_5_categories import LVIS_CATEGORIES as LVIS_V0_5_CATEGORIES
+from detectron2.data.datasets.lvis_v1_categories import LVIS_CATEGORIES as LVIS_V1_CATEGORIES
+from detectron2.structures import BoxMode
+from detectron2.utils.file_io import PathManager
+from fvcore.common.timer import Timer
+
+from .lvis_v1_coco_category_image_count import LVIS_V1_COCO_CATEGORY_IMAGE_COUNT
+
+"""
+This file contains functions to parse LVIS-format annotations into dicts in the
+"Detectron2 format".
+"""
+
+logger = logging.getLogger(__name__)
+
+__all__ = ["custom_load_lvis_json", "custom_register_lvis_instances"]
+
+
+def custom_register_lvis_instances(name, metadata, json_file, image_root):
+    """
+    Register a dataset in LVIS's json annotation format for instance detection and segmentation.
+
+    Args:
+        name (str): a name that identifies the dataset, e.g. "lvis_v0.5_train".
+        metadata (dict): extra metadata associated with this dataset. It can be an empty dict.
+        json_file (str): path to the json instance annotation file.
+        image_root (str or path-like): directory which contains all the images.
+    """
+    DatasetCatalog.register(name, lambda: custom_load_lvis_json(json_file, image_root, name))
+    MetadataCatalog.get(name).set(
+        json_file=json_file, image_root=image_root, evaluator_type="lvis", **metadata
+    )
+
+
+def custom_load_lvis_json(json_file, image_root, dataset_name=None, extra_annotation_keys=None):
+    """
+    Load a json file in LVIS's annotation format.
+
+    Args:
+        json_file (str): full path to the LVIS json annotation file.
+        image_root (str): the directory where the images in this json file exists.
+        dataset_name (str): the name of the dataset (e.g., "lvis_v0.5_train").
+            If provided, this function will put "thing_classes" into the metadata
+            associated with this dataset.
+        extra_annotation_keys (list[str]): list of per-annotation keys that should also be
+            loaded into the dataset dict (besides "bbox", "bbox_mode", "category_id",
+            "segmentation"). The values for these keys will be returned as-is.
+
+    Returns:
+        list[dict]: a list of dicts in Detectron2 standard format. (See
+        `Using Custom Datasets </tutorials/datasets.html>`_ )
+
+    Notes:
+        1. This function does not read the image files.
+           The results do not have the "image" field.
+    """
+    from lvis import LVIS
+
+    json_file = PathManager.get_local_path(json_file)
+
+    timer = Timer()
+    lvis_api = LVIS(json_file)
+    if timer.seconds() > 1:
+        logger.info("Loading {} takes {:.2f} seconds.".format(json_file, timer.seconds()))
+
+    if dataset_name is not None:
+        meta = get_lvis_instances_meta(dataset_name)
+        MetadataCatalog.get(dataset_name).set(**meta)
+
+    # sort indices for reproducible results
+    img_ids = sorted(lvis_api.imgs.keys())
+    # imgs is a list of dicts, each looks something like:
+    # {'license': 4,
+    #  'url': 'http://farm6.staticflickr.com/5454/9413846304_881d5e5c3b_z.jpg',
+    #  'file_name': 'COCO_val2014_000000001268.jpg',
+    #  'height': 427,
+    #  'width': 640,
+    #  'date_captured': '2013-11-17 05:57:24',
+    #  'id': 1268}
+    imgs = lvis_api.load_imgs(img_ids)
+    # anns is a list[list[dict]], where each dict is an annotation
+    # record for an object. The inner list enumerates the objects in an image
+    # and the outer list enumerates over images. Example of anns[0]:
+    # [{'segmentation': [[192.81,
+    #     247.09,
+    #     ...
+    #     219.03,
+    #     249.06]],
+    #   'area': 1035.749,
+    #   'image_id': 1268,
+    #   'bbox': [192.81, 224.8, 74.73, 33.43],
+    #   'category_id': 16,
+    #   'id': 42986},
+    #  ...]
+    anns = [lvis_api.img_ann_map[img_id] for img_id in img_ids]
+
+    # Sanity check that each annotation has a unique id
+    ann_ids = [ann["id"] for anns_per_image in anns for ann in anns_per_image]
+    assert len(set(ann_ids)) == len(ann_ids), "Annotation ids in '{}' are not unique".format(
+        json_file
+    )
+
+    imgs_anns = list(zip(imgs, anns))
+
+    logger.info("Loaded {} images in the LVIS format from {}".format(len(imgs_anns), json_file))
+
+    if extra_annotation_keys:
+        logger.info(
+            "The following extra annotation keys will be loaded: {} ".format(extra_annotation_keys)
+        )
+    else:
+        extra_annotation_keys = []
+
+    def get_file_name(img_root, img_dict):
+        # Determine the path including the split folder ("train2017", "val2017", "test2017") from
+        # the coco_url field. Example:
+        #   'coco_url': 'http://images.cocodataset.org/train2017/000000155379.jpg'
+        if "file_name" in img_dict:
+            file_name = img_dict["file_name"]
+            if img_dict["file_name"].startswith("COCO"):
+                file_name = file_name[-16:]
+            return os.path.join(image_root, file_name)
+        split_folder, file_name = img_dict["coco_url"].split("/")[-2:]
+        return os.path.join(img_root + split_folder, file_name)
+
+    dataset_dicts = []
+
+    for (img_dict, anno_dict_list) in imgs_anns:
+        record = {}
+        record["file_name"] = get_file_name(image_root, img_dict)
+        record["height"] = img_dict["height"]
+        record["width"] = img_dict["width"]
+        record["not_exhaustive_category_ids"] = img_dict.get("not_exhaustive_category_ids", [])
+        record["neg_category_ids"] = img_dict.get("neg_category_ids", [])
+        record["pos_category_ids"] = img_dict.get("pos_category_ids", [])
+        if dataset_name is not None and "thing_dataset_id_to_contiguous_id" in meta:
+            record["neg_category_ids"] = [
+                meta["thing_dataset_id_to_contiguous_id"][x] for x in record["neg_category_ids"]
+            ]
+            record["pos_category_ids"] = [
+                meta["thing_dataset_id_to_contiguous_id"][x] for x in record["pos_category_ids"]
+            ]
+        else:
+            record["neg_category_ids"] = [x - 1 for x in record["neg_category_ids"]]
+            record["pos_category_ids"] = [x - 1 for x in record["pos_category_ids"]]
+        if "captions" in img_dict:
+            record["captions"] = img_dict["captions"]
+        if "caption_features" in img_dict:
+            record["caption_features"] = img_dict["caption_features"]
+        image_id = record["image_id"] = img_dict["id"]
+
+        objs = []
+        for anno in anno_dict_list:
+            assert anno["image_id"] == image_id
+            if anno.get("iscrowd", 0) > 0:
+                continue
+            # Check that the image_id in this annotation is the same as
+            # the image_id we're looking at.
+            # This fails only when the data parsing logic or the annotation file is buggy.
+            assert anno["image_id"] == image_id
+            obj = {"bbox": anno["bbox"], "bbox_mode": BoxMode.XYWH_ABS}
+            # LVIS data loader can be used to load COCO dataset categories. In this case `meta`
+            # variable will have a field with COCO-specific category mapping.
+            if dataset_name is not None and "thing_dataset_id_to_contiguous_id" in meta:
+                obj["category_id"] = meta["thing_dataset_id_to_contiguous_id"][anno["category_id"]]
+            else:
+                obj["category_id"] = anno["category_id"] - 1  # Convert 1-indexed to 0-indexed
+            # segm = anno["segmentation"]  # list[list[float]]
+            # # filter out invalid polygons (< 3 points)
+            # valid_segm = [poly for poly in segm if len(poly) % 2 == 0 and len(poly) >= 6]
+            # assert len(segm) == len(
+            #     valid_segm
+            # ), "Annotation contains an invalid polygon with < 3 points"
+            # assert len(segm) > 0
+            # obj["segmentation"] = segm
+            segm = anno.get("segmentation", None)
+            if segm:  # either list[list[float]] or dict(RLE)
+                if isinstance(segm, dict):
+                    if isinstance(segm["counts"], list):
+                        # convert to compressed RLE
+                        segm = mask_util.frPyObjects(segm, *segm["size"])
+                else:
+                    # filter out invalid polygons (< 3 points)
+                    segm = [poly for poly in segm if len(poly) % 2 == 0 and len(poly) >= 6]
+                    if len(segm) == 0:
+                        num_instances_without_valid_segmentation += 1
+                        continue  # ignore this instance
+                obj["segmentation"] = segm
+
+            phrase = anno.get("phrase", None)
+            if phrase:
+                obj["phrase"] = phrase
+
+            for extra_ann_key in extra_annotation_keys:
+                obj[extra_ann_key] = anno[extra_ann_key]
+            objs.append(obj)
+        record["annotations"] = objs
+        dataset_dicts.append(record)
+
+    return dataset_dicts
+
+
+def get_lvis_instances_meta(dataset_name):
+    """
+    Load LVIS metadata.
+
+    Args:
+        dataset_name (str): LVIS dataset name without the split name (e.g., "lvis_v0.5").
+
+    Returns:
+        dict: LVIS metadata with keys: thing_classes
+    """
+    if "cocofied" in dataset_name:
+        return _get_coco_instances_meta()
+    if "v0.5" in dataset_name:
+        return _get_lvis_instances_meta_v0_5()
+    elif "v1" in dataset_name:
+        return _get_lvis_instances_meta_v1()
+    logger.info("No built-in metadata for dataset {}".format(dataset_name))
+    return {}
+    raise ValueError("No built-in metadata for dataset {}".format(dataset_name))
+
+
+def _get_lvis_instances_meta_v0_5():
+    assert len(LVIS_V0_5_CATEGORIES) == 1230
+    cat_ids = [k["id"] for k in LVIS_V0_5_CATEGORIES]
+    assert min(cat_ids) == 1 and max(cat_ids) == len(
+        cat_ids
+    ), "Category ids are not in [1, #categories], as expected"
+    # Ensure that the category list is sorted by id
+    lvis_categories = sorted(LVIS_V0_5_CATEGORIES, key=lambda x: x["id"])
+    thing_classes = [k["synonyms"][0] for k in lvis_categories]
+    meta = {"thing_classes": thing_classes}
+    return meta
+
+
+def _get_lvis_instances_meta_v1():
+    assert len(LVIS_V1_CATEGORIES) == 1203
+    cat_ids = [k["id"] for k in LVIS_V1_CATEGORIES]
+    assert min(cat_ids) == 1 and max(cat_ids) == len(
+        cat_ids
+    ), "Category ids are not in [1, #categories], as expected"
+    # Ensure that the category list is sorted by id
+    lvis_categories = sorted(LVIS_V1_CATEGORIES, key=lambda x: x["id"])
+    thing_classes = [k["synonyms"][0] for k in lvis_categories]
+    meta = {"thing_classes": thing_classes, "class_image_count": LVIS_V1_COCO_CATEGORY_IMAGE_COUNT}
+    return meta
+
+
+_PREDEFINED_SPLITS_LVIS = {
+    "lvis_v1_train+coco": {
+        "lvis_v1_train+coco": ("coco/", "lvis/lvis_v1_train+coco_mask.json"),
+    },
+    "lvis_v1_val+coco": {
+        "lvis_v1_val+coco": ("coco/", "lvis/lvis_v1_val+coco_mask.json"),
+    },
+    "lvis_v1_minival": {
+        "lvis_v1_minival": ("coco/", "lvis/lvis_v1_minival_inserted_image_name.json"),
+    },
+}
+
+
+def register_all_lvis_coco(root):
+    for dataset_name, splits_per_dataset in _PREDEFINED_SPLITS_LVIS.items():
+        for key, (image_root, json_file) in splits_per_dataset.items():
+            custom_register_lvis_instances(
+                key,
+                get_lvis_instances_meta(dataset_name),
+                os.path.join(root, json_file) if "://" not in json_file else json_file,
+                os.path.join(root, image_root),
+            )
+
+
+if __name__.endswith(".lvis_coco"):
+    # Assume pre-defined datasets live in `./datasets`.
+    _root = os.getenv("DETECTRON2_DATASETS", "datasets")
+    register_all_lvis_coco(_root)

approach/ovod/APE/ape/data/datasets/lvis_coco_panoptic.py

@@ -0,0 +1,192 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+import copy
+import json
+import os
+
+from detectron2.data import DatasetCatalog, MetadataCatalog
+from detectron2.data.datasets.builtin_meta import COCO_CATEGORIES
+from detectron2.data.datasets.coco import load_sem_seg
+from detectron2.utils.file_io import PathManager
+
+from .lvis_coco import custom_load_lvis_json, get_lvis_instances_meta
+
+__all__ = ["register_lvis_panoptic_separated"]
+
+
+def register_lvis_panoptic_separated(
+    name, metadata, image_root, panoptic_root, panoptic_json, sem_seg_root, instances_json
+):
+    """
+    Register a "separated" version of COCO panoptic segmentation dataset named `name`.
+    The annotations in this registered dataset will contain both instance annotations and
+    semantic annotations, each with its own contiguous ids. Hence it's called "separated".
+
+    It follows the setting used by the PanopticFPN paper:
+
+    1. The instance annotations directly come from polygons in the COCO
+       instances annotation task, rather than from the masks in the COCO panoptic annotations.
+
+       The two format have small differences:
+       Polygons in the instance annotations may have overlaps.
+       The mask annotations are produced by labeling the overlapped polygons
+       with depth ordering.
+
+    2. The semantic annotations are converted from panoptic annotations, where
+       all "things" are assigned a semantic id of 0.
+       All semantic categories will therefore have ids in contiguous
+       range [1, #stuff_categories].
+
+    This function will also register a pure semantic segmentation dataset
+    named ``name + '_stuffonly'``.
+
+    Args:
+        name (str): the name that identifies a dataset,
+            e.g. "coco_2017_train_panoptic"
+        metadata (dict): extra metadata associated with this dataset.
+        image_root (str): directory which contains all the images
+        panoptic_root (str): directory which contains panoptic annotation images
+        panoptic_json (str): path to the json panoptic annotation file
+        sem_seg_root (str): directory which contains all the ground truth segmentation annotations.
+        instances_json (str): path to the json instance annotation file
+    """
+    panoptic_name = name + "_separated"
+    split_folder = sem_seg_root.split("_")[-1]  # datasets/coco/panoptic_stuff_train2017
+    DatasetCatalog.register(
+        panoptic_name,
+        lambda: merge_to_panoptic(
+            custom_load_lvis_json(instances_json, image_root, panoptic_name),
+            load_sem_seg(sem_seg_root, os.path.join(image_root, split_folder)),
+        ),
+    )
+    MetadataCatalog.get(panoptic_name).set(
+        panoptic_root=panoptic_root,
+        image_root=image_root,
+        panoptic_json=panoptic_json,
+        sem_seg_root=sem_seg_root,
+        json_file=instances_json,  # TODO rename
+        evaluator_type="coco_panoptic_seg",
+        ignore_label=255,
+        **metadata,
+    )
+
+    semantic_name = name + "_stuffonly"
+    DatasetCatalog.register(semantic_name, lambda: load_sem_seg(sem_seg_root, image_root))
+    MetadataCatalog.get(semantic_name).set(
+        sem_seg_root=sem_seg_root,
+        image_root=image_root,
+        evaluator_type="sem_seg",
+        ignore_label=255,
+        **metadata,
+    )
+
+
+def merge_to_panoptic(detection_dicts, sem_seg_dicts):
+    """
+    Create dataset dicts for panoptic segmentation, by
+    merging two dicts using "file_name" field to match their entries.
+
+    Args:
+        detection_dicts (list[dict]): lists of dicts for object detection or instance segmentation.
+        sem_seg_dicts (list[dict]): lists of dicts for semantic segmentation.
+
+    Returns:
+        list[dict] (one per input image): Each dict contains all (key, value) pairs from dicts in
+            both detection_dicts and sem_seg_dicts that correspond to the same image.
+            The function assumes that the same key in different dicts has the same value.
+    """
+    results = []
+    sem_seg_file_to_entry = {x["file_name"]: x for x in sem_seg_dicts}
+    assert len(sem_seg_file_to_entry) > 0
+
+    for det_dict in detection_dicts:
+        dic = copy.copy(det_dict)
+        dic.update(sem_seg_file_to_entry[dic["file_name"]])
+        results.append(dic)
+    return results
+
+
+def _get_builtin_metadata(dataset_name):
+    if dataset_name == "lvis_panoptic_separated":
+        return _get_lvis_panoptic_separated_meta()
+
+    raise KeyError("No built-in metadata for dataset {}".format(dataset_name))
+
+
+def _get_lvis_panoptic_separated_meta():
+    """
+    Returns metadata for "separated" version of the panoptic segmentation dataset.
+    """
+    stuff_ids = [k["id"] for k in COCO_CATEGORIES if k["isthing"] == 0]
+    assert len(stuff_ids) == 53, len(stuff_ids)
+
+    # For semantic segmentation, this mapping maps from contiguous stuff id
+    # (in [0, 53], used in models) to ids in the dataset (used for processing results)
+    # The id 0 is mapped to an extra category "thing".
+    stuff_dataset_id_to_contiguous_id = {k: i + 1 for i, k in enumerate(stuff_ids)}
+    # When converting COCO panoptic annotations to semantic annotations
+    # We label the "thing" category to 0
+    stuff_dataset_id_to_contiguous_id[0] = 0
+
+    # 54 names for COCO stuff categories (including "things")
+    stuff_classes = ["things"] + [
+        k["name"].replace("-other", "").replace("-merged", "").replace("-stuff", "")
+        for k in COCO_CATEGORIES
+        if k["isthing"] == 0
+    ]
+
+    # NOTE: I randomly picked a color for things
+    stuff_colors = [[82, 18, 128]] + [k["color"] for k in COCO_CATEGORIES if k["isthing"] == 0]
+    ret = {
+        "stuff_dataset_id_to_contiguous_id": stuff_dataset_id_to_contiguous_id,
+        "stuff_classes": stuff_classes,
+        "stuff_colors": stuff_colors,
+    }
+    ret.update(get_lvis_instances_meta("v1"))
+    return ret
+
+
+_PREDEFINED_SPLITS_LVIS_PANOPTIC = {
+    "lvis_v1_train+coco_panoptic": (
+        # This is the original panoptic annotation directory
+        "coco/panoptic_train2017",
+        "coco/annotations/panoptic_train2017.json",
+        # This directory contains semantic annotations that are
+        # converted from panoptic annotations.
+        # It is used by PanopticFPN.
+        # You can use the script at detectron2/datasets/prepare_panoptic_fpn.py
+        # to create these directories.
+        "coco/panoptic_stuff_train2017",
+    ),
+    "lvis_v1_val+coco_panoptic": (
+        "coco/panoptic_val2017",
+        "coco/annotations/panoptic_val2017.json",
+        "coco/panoptic_stuff_val2017",
+    ),
+}
+
+
+def register_all_lvis_coco_panoptic(root):
+    for (
+        prefix,
+        (panoptic_root, panoptic_json, semantic_root),
+    ) in _PREDEFINED_SPLITS_LVIS_PANOPTIC.items():
+        prefix_instances = prefix[: -len("_panoptic")]
+        instances_meta = MetadataCatalog.get(prefix_instances)
+        image_root, instances_json = instances_meta.image_root, instances_meta.json_file
+        # The "separated" version of COCO panoptic segmentation dataset,
+        # e.g. used by Panoptic FPN
+        register_lvis_panoptic_separated(
+            prefix,
+            _get_builtin_metadata("lvis_panoptic_separated"),
+            image_root,
+            os.path.join(root, panoptic_root),
+            os.path.join(root, panoptic_json),
+            os.path.join(root, semantic_root),
+            instances_json,
+        )
+
+
+if __name__.endswith(".lvis_coco_panoptic"):
+    # Assume pre-defined datasets live in `./datasets`.
+    _root = os.getenv("DETECTRON2_DATASETS", "datasets")
+    register_all_lvis_coco_panoptic(_root)

approach/ovod/APE/ape/data/datasets/lvis_v1_coco_category_image_count.py

@@ -0,0 +1,20 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+# Autogen with
+# with open("lvis_v1_train.json", "r") as f:
+#     a = json.load(f)
+# c = a["categories"]
+# for x in c:
+# del x["name"]
+# del x["instance_count"]
+# del x["def"]
+# del x["synonyms"]
+# del x["frequency"]
+# del x["synset"]
+# LVIS_CATEGORY_IMAGE_COUNT = repr(c) + "  # noqa"
+# with open("/tmp/lvis_category_image_count.py", "wt") as f:
+#     f.write(f"LVIS_CATEGORY_IMAGE_COUNT = {LVIS_CATEGORY_IMAGE_COUNT}")
+# Then paste the contents of that file below
+
+# fmt: off
+LVIS_V1_COCO_CATEGORY_IMAGE_COUNT = [{'image_count': 64, 'id': 1}, {'image_count': 364, 'id': 2}, {'image_count': 2596, 'id': 3}, {'image_count': 149, 'id': 4}, {'image_count': 29, 'id': 5}, {'image_count': 26, 'id': 6}, {'image_count': 59, 'id': 7}, {'image_count': 22, 'id': 8}, {'image_count': 12, 'id': 9}, {'image_count': 28, 'id': 10}, {'image_count': 505, 'id': 11}, {'image_count': 1626, 'id': 12}, {'image_count': 4, 'id': 13}, {'image_count': 10, 'id': 14}, {'image_count': 500, 'id': 15}, {'image_count': 33, 'id': 16}, {'image_count': 3, 'id': 17}, {'image_count': 44, 'id': 18}, {'image_count': 561, 'id': 19}, {'image_count': 8, 'id': 20}, {'image_count': 9, 'id': 21}, {'image_count': 33, 'id': 22}, {'image_count': 1883, 'id': 23}, {'image_count': 98, 'id': 24}, {'image_count': 70, 'id': 25}, {'image_count': 46, 'id': 26}, {'image_count': 117, 'id': 27}, {'image_count': 41, 'id': 28}, {'image_count': 1395, 'id': 29}, {'image_count': 7, 'id': 30}, {'image_count': 1, 'id': 31}, {'image_count': 314, 'id': 32}, {'image_count': 31, 'id': 33}, {'image_count': 4965, 'id': 34}, {'image_count': 6090, 'id': 35}, {'image_count': 2183, 'id': 36}, {'image_count': 47, 'id': 37}, {'image_count': 3, 'id': 38}, {'image_count': 3, 'id': 39}, {'image_count': 1, 'id': 40}, {'image_count': 3780, 'id': 41}, {'image_count': 6, 'id': 42}, {'image_count': 210, 'id': 43}, {'image_count': 36, 'id': 44}, {'image_count': 2000, 'id': 45}, {'image_count': 17, 'id': 46}, {'image_count': 51, 'id': 47}, {'image_count': 138, 'id': 48}, {'image_count': 3, 'id': 49}, {'image_count': 1470, 'id': 50}, {'image_count': 3, 'id': 51}, {'image_count': 2, 'id': 52}, {'image_count': 186, 'id': 53}, {'image_count': 76, 'id': 54}, {'image_count': 26, 'id': 55}, {'image_count': 303, 'id': 56}, {'image_count': 738, 'id': 57}, {'image_count': 2216, 'id': 58}, {'image_count': 1934, 'id': 59}, {'image_count': 2287, 'id': 60}, {'image_count': 1622, 'id': 61}, {'image_count': 41, 'id': 62}, {'image_count': 4, 'id': 63}, {'image_count': 11, 'id': 64}, {'image_count': 270, 'id': 65}, {'image_count': 349, 'id': 66}, {'image_count': 42, 'id': 67}, {'image_count': 823, 'id': 68}, {'image_count': 6, 'id': 69}, {'image_count': 48, 'id': 70}, {'image_count': 3, 'id': 71}, {'image_count': 42, 'id': 72}, {'image_count': 24, 'id': 73}, {'image_count': 16, 'id': 74}, {'image_count': 605, 'id': 75}, {'image_count': 898, 'id': 76}, {'image_count': 3194, 'id': 77}, {'image_count': 2, 'id': 78}, {'image_count': 125, 'id': 79}, {'image_count': 1739, 'id': 80}, {'image_count': 140, 'id': 81}, {'image_count': 4, 'id': 82}, {'image_count': 322, 'id': 83}, {'image_count': 60, 'id': 84}, {'image_count': 2, 'id': 85}, {'image_count': 231, 'id': 86}, {'image_count': 333, 'id': 87}, {'image_count': 1941, 'id': 88}, {'image_count': 367, 'id': 89}, {'image_count': 4908, 'id': 90}, {'image_count': 18, 'id': 91}, {'image_count': 81, 'id': 92}, {'image_count': 1, 'id': 93}, {'image_count': 2940, 'id': 94}, {'image_count': 430, 'id': 95}, {'image_count': 247, 'id': 96}, {'image_count': 94, 'id': 97}, {'image_count': 21, 'id': 98}, {'image_count': 2950, 'id': 99}, {'image_count': 16, 'id': 100}, {'image_count': 12, 'id': 101}, {'image_count': 25, 'id': 102}, {'image_count': 41, 'id': 103}, {'image_count': 244, 'id': 104}, {'image_count': 7, 'id': 105}, {'image_count': 1, 'id': 106}, {'image_count': 40, 'id': 107}, {'image_count': 40, 'id': 108}, {'image_count': 104, 'id': 109}, {'image_count': 1671, 'id': 110}, {'image_count': 49, 'id': 111}, {'image_count': 243, 'id': 112}, {'image_count': 2, 'id': 113}, {'image_count': 242, 'id': 114}, {'image_count': 271, 'id': 115}, {'image_count': 104, 'id': 116}, {'image_count': 8, 'id': 117}, {'image_count': 2674, 'id': 118}, {'image_count': 1, 'id': 119}, {'image_count': 48, 'id': 120}, {'image_count': 14, 'id': 121}, {'image_count': 40, 'id': 122}, {'image_count': 1, 'id': 123}, {'image_count': 37, 'id': 124}, {'image_count': 1510, 'id': 125}, {'image_count': 6, 'id': 126}, {'image_count': 4808, 'id': 127}, {'image_count': 70, 'id': 128}, {'image_count': 86, 'id': 129}, {'image_count': 7, 'id': 130}, {'image_count': 5, 'id': 131}, {'image_count': 1406, 'id': 132}, {'image_count': 7655, 'id': 133}, {'image_count': 15, 'id': 134}, {'image_count': 28, 'id': 135}, {'image_count': 6, 'id': 136}, {'image_count': 494, 'id': 137}, {'image_count': 234, 'id': 138}, {'image_count': 6332, 'id': 139}, {'image_count': 1, 'id': 140}, {'image_count': 35, 'id': 141}, {'image_count': 5, 'id': 142}, {'image_count': 1828, 'id': 143}, {'image_count': 8, 'id': 144}, {'image_count': 63, 'id': 145}, {'image_count': 1668, 'id': 146}, {'image_count': 4, 'id': 147}, {'image_count': 95, 'id': 148}, {'image_count': 17, 'id': 149}, {'image_count': 1567, 'id': 150}, {'image_count': 2, 'id': 151}, {'image_count': 103, 'id': 152}, {'image_count': 50, 'id': 153}, {'image_count': 1678, 'id': 154}, {'image_count': 6, 'id': 155}, {'image_count': 92, 'id': 156}, {'image_count': 19, 'id': 157}, {'image_count': 37, 'id': 158}, {'image_count': 4, 'id': 159}, {'image_count': 709, 'id': 160}, {'image_count': 9, 'id': 161}, {'image_count': 82, 'id': 162}, {'image_count': 15, 'id': 163}, {'image_count': 3, 'id': 164}, {'image_count': 61, 'id': 165}, {'image_count': 51, 'id': 166}, {'image_count': 5, 'id': 167}, {'image_count': 13, 'id': 168}, {'image_count': 642, 'id': 169}, {'image_count': 24, 'id': 170}, {'image_count': 255, 'id': 171}, {'image_count': 9, 'id': 172}, {'image_count': 3431, 'id': 173}, {'image_count': 31, 'id': 174}, {'image_count': 158, 'id': 175}, {'image_count': 80, 'id': 176}, {'image_count': 1884, 'id': 177}, {'image_count': 158, 'id': 178}, {'image_count': 2, 'id': 179}, {'image_count': 12, 'id': 180}, {'image_count': 1659, 'id': 181}, {'image_count': 7, 'id': 182}, {'image_count': 2555, 'id': 183}, {'image_count': 57, 'id': 184}, {'image_count': 174, 'id': 185}, {'image_count': 95, 'id': 186}, {'image_count': 27, 'id': 187}, {'image_count': 22, 'id': 188}, {'image_count': 1391, 'id': 189}, {'image_count': 90, 'id': 190}, {'image_count': 40, 'id': 191}, {'image_count': 445, 'id': 192}, {'image_count': 21, 'id': 193}, {'image_count': 1132, 'id': 194}, {'image_count': 177, 'id': 195}, {'image_count': 4, 'id': 196}, {'image_count': 17, 'id': 197}, {'image_count': 84, 'id': 198}, {'image_count': 55, 'id': 199}, {'image_count': 30, 'id': 200}, {'image_count': 25, 'id': 201}, {'image_count': 2, 'id': 202}, {'image_count': 125, 'id': 203}, {'image_count': 1135, 'id': 204}, {'image_count': 19, 'id': 205}, {'image_count': 72, 'id': 206}, {'image_count': 10661, 'id': 207}, {'image_count': 159, 'id': 208}, {'image_count': 7, 'id': 209}, {'image_count': 1, 'id': 210}, {'image_count': 13, 'id': 211}, {'image_count': 35, 'id': 212}, {'image_count': 18, 'id': 213}, {'image_count': 8, 'id': 214}, {'image_count': 6, 'id': 215}, {'image_count': 35, 'id': 216}, {'image_count': 1531, 'id': 217}, {'image_count': 103, 'id': 218}, {'image_count': 28, 'id': 219}, {'image_count': 63, 'id': 220}, {'image_count': 28, 'id': 221}, {'image_count': 5, 'id': 222}, {'image_count': 7, 'id': 223}, {'image_count': 14, 'id': 224}, {'image_count': 3422, 'id': 225}, {'image_count': 133, 'id': 226}, {'image_count': 16, 'id': 227}, {'image_count': 27, 'id': 228}, {'image_count': 110, 'id': 229}, {'image_count': 4201, 'id': 230}, {'image_count': 4, 'id': 231}, {'image_count': 11104, 'id': 232}, {'image_count': 8, 'id': 233}, {'image_count': 1, 'id': 234}, {'image_count': 263, 'id': 235}, {'image_count': 10, 'id': 236}, {'image_count': 2, 'id': 237}, {'image_count': 3, 'id': 238}, {'image_count': 87, 'id': 239}, {'image_count': 9, 'id': 240}, {'image_count': 71, 'id': 241}, {'image_count': 13, 'id': 242}, {'image_count': 18, 'id': 243}, {'image_count': 2, 'id': 244}, {'image_count': 5, 'id': 245}, {'image_count': 45, 'id': 246}, {'image_count': 1, 'id': 247}, {'image_count': 23, 'id': 248}, {'image_count': 32, 'id': 249}, {'image_count': 4, 'id': 250}, {'image_count': 1, 'id': 251}, {'image_count': 858, 'id': 252}, {'image_count': 661, 'id': 253}, {'image_count': 168, 'id': 254}, {'image_count': 210, 'id': 255}, {'image_count': 65, 'id': 256}, {'image_count': 4, 'id': 257}, {'image_count': 2, 'id': 258}, {'image_count': 159, 'id': 259}, {'image_count': 31, 'id': 260}, {'image_count': 811, 'id': 261}, {'image_count': 1, 'id': 262}, {'image_count': 42, 'id': 263}, {'image_count': 27, 'id': 264}, {'image_count': 2, 'id': 265}, {'image_count': 5, 'id': 266}, {'image_count': 95, 'id': 267}, {'image_count': 32, 'id': 268}, {'image_count': 1, 'id': 269}, {'image_count': 1, 'id': 270}, {'image_count': 4053, 'id': 271}, {'image_count': 897, 'id': 272}, {'image_count': 31, 'id': 273}, {'image_count': 23, 'id': 274}, {'image_count': 1, 'id': 275}, {'image_count': 202, 'id': 276}, {'image_count': 746, 'id': 277}, {'image_count': 44, 'id': 278}, {'image_count': 14, 'id': 279}, {'image_count': 26, 'id': 280}, {'image_count': 1, 'id': 281}, {'image_count': 2, 'id': 282}, {'image_count': 25, 'id': 283}, {'image_count': 238, 'id': 284}, {'image_count': 592, 'id': 285}, {'image_count': 26, 'id': 286}, {'image_count': 5, 'id': 287}, {'image_count': 42, 'id': 288}, {'image_count': 13, 'id': 289}, {'image_count': 46, 'id': 290}, {'image_count': 1, 'id': 291}, {'image_count': 8, 'id': 292}, {'image_count': 34, 'id': 293}, {'image_count': 5, 'id': 294}, {'image_count': 1, 'id': 295}, {'image_count': 2356, 'id': 296}, {'image_count': 717, 'id': 297}, {'image_count': 1010, 'id': 298}, {'image_count': 679, 'id': 299}, {'image_count': 3, 'id': 300}, {'image_count': 4, 'id': 301}, {'image_count': 1, 'id': 302}, {'image_count': 166, 'id': 303}, {'image_count': 2, 'id': 304}, {'image_count': 266, 'id': 305}, {'image_count': 101, 'id': 306}, {'image_count': 6, 'id': 307}, {'image_count': 14, 'id': 308}, {'image_count': 133, 'id': 309}, {'image_count': 2, 'id': 310}, {'image_count': 38, 'id': 311}, {'image_count': 95, 'id': 312}, {'image_count': 1, 'id': 313}, {'image_count': 12, 'id': 314}, {'image_count': 49, 'id': 315}, {'image_count': 5, 'id': 316}, {'image_count': 5, 'id': 317}, {'image_count': 16, 'id': 318}, {'image_count': 216, 'id': 319}, {'image_count': 12, 'id': 320}, {'image_count': 1, 'id': 321}, {'image_count': 54, 'id': 322}, {'image_count': 5, 'id': 323}, {'image_count': 245, 'id': 324}, {'image_count': 12, 'id': 325}, {'image_count': 7, 'id': 326}, {'image_count': 35, 'id': 327}, {'image_count': 36, 'id': 328}, {'image_count': 32, 'id': 329}, {'image_count': 1027, 'id': 330}, {'image_count': 10, 'id': 331}, {'image_count': 12, 'id': 332}, {'image_count': 1, 'id': 333}, {'image_count': 67, 'id': 334}, {'image_count': 71, 'id': 335}, {'image_count': 30, 'id': 336}, {'image_count': 48, 'id': 337}, {'image_count': 249, 'id': 338}, {'image_count': 13, 'id': 339}, {'image_count': 29, 'id': 340}, {'image_count': 14, 'id': 341}, {'image_count': 236, 'id': 342}, {'image_count': 15, 'id': 343}, {'image_count': 8183, 'id': 344}, {'image_count': 25, 'id': 345}, {'image_count': 249, 'id': 346}, {'image_count': 139, 'id': 347}, {'image_count': 2, 'id': 348}, {'image_count': 2, 'id': 349}, {'image_count': 1890, 'id': 350}, {'image_count': 1240, 'id': 351}, {'image_count': 1, 'id': 352}, {'image_count': 9, 'id': 353}, {'image_count': 1, 'id': 354}, {'image_count': 3, 'id': 355}, {'image_count': 11, 'id': 356}, {'image_count': 4, 'id': 357}, {'image_count': 236, 'id': 358}, {'image_count': 44, 'id': 359}, {'image_count': 19, 'id': 360}, {'image_count': 1100, 'id': 361}, {'image_count': 7, 'id': 362}, {'image_count': 69, 'id': 363}, {'image_count': 2, 'id': 364}, {'image_count': 8, 'id': 365}, {'image_count': 5, 'id': 366}, {'image_count': 10225, 'id': 367}, {'image_count': 6, 'id': 368}, {'image_count': 106, 'id': 369}, {'image_count': 81, 'id': 370}, {'image_count': 17, 'id': 371}, {'image_count': 134, 'id': 372}, {'image_count': 312, 'id': 373}, {'image_count': 8, 'id': 374}, {'image_count': 271, 'id': 375}, {'image_count': 2, 'id': 376}, {'image_count': 103, 'id': 377}, {'image_count': 3758, 'id': 378}, {'image_count': 574, 'id': 379}, {'image_count': 120, 'id': 380}, {'image_count': 2, 'id': 381}, {'image_count': 2, 'id': 382}, {'image_count': 13, 'id': 383}, {'image_count': 29, 'id': 384}, {'image_count': 1710, 'id': 385}, {'image_count': 66, 'id': 386}, {'image_count': 1326, 'id': 387}, {'image_count': 1, 'id': 388}, {'image_count': 3, 'id': 389}, {'image_count': 1942, 'id': 390}, {'image_count': 19, 'id': 391}, {'image_count': 1488, 'id': 392}, {'image_count': 46, 'id': 393}, {'image_count': 106, 'id': 394}, {'image_count': 115, 'id': 395}, {'image_count': 19, 'id': 396}, {'image_count': 2, 'id': 397}, {'image_count': 1, 'id': 398}, {'image_count': 28, 'id': 399}, {'image_count': 9, 'id': 400}, {'image_count': 192, 'id': 401}, {'image_count': 12, 'id': 402}, {'image_count': 21, 'id': 403}, {'image_count': 247, 'id': 404}, {'image_count': 6, 'id': 405}, {'image_count': 64, 'id': 406}, {'image_count': 7, 'id': 407}, {'image_count': 40, 'id': 408}, {'image_count': 542, 'id': 409}, {'image_count': 2, 'id': 410}, {'image_count': 1898, 'id': 411}, {'image_count': 36, 'id': 412}, {'image_count': 4, 'id': 413}, {'image_count': 1, 'id': 414}, {'image_count': 191, 'id': 415}, {'image_count': 6, 'id': 416}, {'image_count': 41, 'id': 417}, {'image_count': 39, 'id': 418}, {'image_count': 46, 'id': 419}, {'image_count': 1, 'id': 420}, {'image_count': 2100, 'id': 421}, {'image_count': 1915, 'id': 422}, {'image_count': 11, 'id': 423}, {'image_count': 82, 'id': 424}, {'image_count': 18, 'id': 425}, {'image_count': 1, 'id': 426}, {'image_count': 7, 'id': 427}, {'image_count': 3, 'id': 428}, {'image_count': 575, 'id': 429}, {'image_count': 1907, 'id': 430}, {'image_count': 8, 'id': 431}, {'image_count': 4, 'id': 432}, {'image_count': 32, 'id': 433}, {'image_count': 11, 'id': 434}, {'image_count': 4, 'id': 435}, {'image_count': 54, 'id': 436}, {'image_count': 202, 'id': 437}, {'image_count': 32, 'id': 438}, {'image_count': 3, 'id': 439}, {'image_count': 130, 'id': 440}, {'image_count': 119, 'id': 441}, {'image_count': 141, 'id': 442}, {'image_count': 29, 'id': 443}, {'image_count': 525, 'id': 444}, {'image_count': 1537, 'id': 445}, {'image_count': 2, 'id': 446}, {'image_count': 113, 'id': 447}, {'image_count': 16, 'id': 448}, {'image_count': 7, 'id': 449}, {'image_count': 35, 'id': 450}, {'image_count': 1908, 'id': 451}, {'image_count': 353, 'id': 452}, {'image_count': 18, 'id': 453}, {'image_count': 14, 'id': 454}, {'image_count': 77, 'id': 455}, {'image_count': 8, 'id': 456}, {'image_count': 37, 'id': 457}, {'image_count': 1, 'id': 458}, {'image_count': 346, 'id': 459}, {'image_count': 19, 'id': 460}, {'image_count': 1779, 'id': 461}, {'image_count': 23, 'id': 462}, {'image_count': 25, 'id': 463}, {'image_count': 67, 'id': 464}, {'image_count': 19, 'id': 465}, {'image_count': 28, 'id': 466}, {'image_count': 4, 'id': 467}, {'image_count': 27, 'id': 468}, {'image_count': 3150, 'id': 469}, {'image_count': 11, 'id': 470}, {'image_count': 13, 'id': 471}, {'image_count': 13, 'id': 472}, {'image_count': 32, 'id': 473}, {'image_count': 1911, 'id': 474}, {'image_count': 42, 'id': 475}, {'image_count': 17, 'id': 476}, {'image_count': 128, 'id': 477}, {'image_count': 1, 'id': 478}, {'image_count': 9, 'id': 479}, {'image_count': 10, 'id': 480}, {'image_count': 4, 'id': 481}, {'image_count': 9, 'id': 482}, {'image_count': 18, 'id': 483}, {'image_count': 41, 'id': 484}, {'image_count': 28, 'id': 485}, {'image_count': 3, 'id': 486}, {'image_count': 65, 'id': 487}, {'image_count': 9, 'id': 488}, {'image_count': 23, 'id': 489}, {'image_count': 24, 'id': 490}, {'image_count': 1, 'id': 491}, {'image_count': 2, 'id': 492}, {'image_count': 59, 'id': 493}, {'image_count': 48, 'id': 494}, {'image_count': 17, 'id': 495}, {'image_count': 2056, 'id': 496}, {'image_count': 18, 'id': 497}, {'image_count': 1920, 'id': 498}, {'image_count': 50, 'id': 499}, {'image_count': 1890, 'id': 500}, {'image_count': 99, 'id': 501}, {'image_count': 1530, 'id': 502}, {'image_count': 3, 'id': 503}, {'image_count': 11, 'id': 504}, {'image_count': 19, 'id': 505}, {'image_count': 3, 'id': 506}, {'image_count': 63, 'id': 507}, {'image_count': 5, 'id': 508}, {'image_count': 6, 'id': 509}, {'image_count': 233, 'id': 510}, {'image_count': 54, 'id': 511}, {'image_count': 36, 'id': 512}, {'image_count': 10, 'id': 513}, {'image_count': 124, 'id': 514}, {'image_count': 101, 'id': 515}, {'image_count': 3, 'id': 516}, {'image_count': 363, 'id': 517}, {'image_count': 3, 'id': 518}, {'image_count': 30, 'id': 519}, {'image_count': 18, 'id': 520}, {'image_count': 199, 'id': 521}, {'image_count': 97, 'id': 522}, {'image_count': 32, 'id': 523}, {'image_count': 121, 'id': 524}, {'image_count': 16, 'id': 525}, {'image_count': 12, 'id': 526}, {'image_count': 2, 'id': 527}, {'image_count': 214, 'id': 528}, {'image_count': 48, 'id': 529}, {'image_count': 26, 'id': 530}, {'image_count': 13, 'id': 531}, {'image_count': 4, 'id': 532}, {'image_count': 11, 'id': 533}, {'image_count': 171, 'id': 534}, {'image_count': 7, 'id': 535}, {'image_count': 200, 'id': 536}, {'image_count': 91, 'id': 537}, {'image_count': 9, 'id': 538}, {'image_count': 72, 'id': 539}, {'image_count': 1886, 'id': 540}, {'image_count': 4, 'id': 541}, {'image_count': 1, 'id': 542}, {'image_count': 1, 'id': 543}, {'image_count': 1932, 'id': 544}, {'image_count': 4, 'id': 545}, {'image_count': 56, 'id': 546}, {'image_count': 854, 'id': 547}, {'image_count': 755, 'id': 548}, {'image_count': 1843, 'id': 549}, {'image_count': 96, 'id': 550}, {'image_count': 7, 'id': 551}, {'image_count': 74, 'id': 552}, {'image_count': 66, 'id': 553}, {'image_count': 57, 'id': 554}, {'image_count': 44, 'id': 555}, {'image_count': 1905, 'id': 556}, {'image_count': 4, 'id': 557}, {'image_count': 90, 'id': 558}, {'image_count': 1635, 'id': 559}, {'image_count': 8, 'id': 560}, {'image_count': 5, 'id': 561}, {'image_count': 50, 'id': 562}, {'image_count': 545, 'id': 563}, {'image_count': 20, 'id': 564}, {'image_count': 193, 'id': 565}, {'image_count': 285, 'id': 566}, {'image_count': 3, 'id': 567}, {'image_count': 1, 'id': 568}, {'image_count': 2488, 'id': 569}, {'image_count': 294, 'id': 570}, {'image_count': 3, 'id': 571}, {'image_count': 5, 'id': 572}, {'image_count': 24, 'id': 573}, {'image_count': 2, 'id': 574}, {'image_count': 2, 'id': 575}, {'image_count': 16, 'id': 576}, {'image_count': 8, 'id': 577}, {'image_count': 154, 'id': 578}, {'image_count': 66, 'id': 579}, {'image_count': 1, 'id': 580}, {'image_count': 24, 'id': 581}, {'image_count': 1, 'id': 582}, {'image_count': 4, 'id': 583}, {'image_count': 75, 'id': 584}, {'image_count': 6, 'id': 585}, {'image_count': 126, 'id': 586}, {'image_count': 24, 'id': 587}, {'image_count': 22, 'id': 588}, {'image_count': 1872, 'id': 589}, {'image_count': 16, 'id': 590}, {'image_count': 423, 'id': 591}, {'image_count': 1927, 'id': 592}, {'image_count': 38, 'id': 593}, {'image_count': 3, 'id': 594}, {'image_count': 1945, 'id': 595}, {'image_count': 35, 'id': 596}, {'image_count': 1, 'id': 597}, {'image_count': 13, 'id': 598}, {'image_count': 9, 'id': 599}, {'image_count': 14, 'id': 600}, {'image_count': 37, 'id': 601}, {'image_count': 3, 'id': 602}, {'image_count': 4, 'id': 603}, {'image_count': 100, 'id': 604}, {'image_count': 195, 'id': 605}, {'image_count': 1, 'id': 606}, {'image_count': 12, 'id': 607}, {'image_count': 24, 'id': 608}, {'image_count': 489, 'id': 609}, {'image_count': 10, 'id': 610}, {'image_count': 1921, 'id': 611}, {'image_count': 42, 'id': 612}, {'image_count': 81, 'id': 613}, {'image_count': 894, 'id': 614}, {'image_count': 3839, 'id': 615}, {'image_count': 7, 'id': 616}, {'image_count': 1567, 'id': 617}, {'image_count': 10, 'id': 618}, {'image_count': 8, 'id': 619}, {'image_count': 7, 'id': 620}, {'image_count': 629, 'id': 621}, {'image_count': 89, 'id': 622}, {'image_count': 15, 'id': 623}, {'image_count': 134, 'id': 624}, {'image_count': 4, 'id': 625}, {'image_count': 1802, 'id': 626}, {'image_count': 595, 'id': 627}, {'image_count': 1210, 'id': 628}, {'image_count': 48, 'id': 629}, {'image_count': 418, 'id': 630}, {'image_count': 3050, 'id': 631}, {'image_count': 5, 'id': 632}, {'image_count': 221, 'id': 633}, {'image_count': 10, 'id': 634}, {'image_count': 7, 'id': 635}, {'image_count': 76, 'id': 636}, {'image_count': 22, 'id': 637}, {'image_count': 10, 'id': 638}, {'image_count': 341, 'id': 639}, {'image_count': 1, 'id': 640}, {'image_count': 705, 'id': 641}, {'image_count': 1900, 'id': 642}, {'image_count': 188, 'id': 643}, {'image_count': 227, 'id': 644}, {'image_count': 861, 'id': 645}, {'image_count': 6, 'id': 646}, {'image_count': 115, 'id': 647}, {'image_count': 5, 'id': 648}, {'image_count': 43, 'id': 649}, {'image_count': 14, 'id': 650}, {'image_count': 6, 'id': 651}, {'image_count': 15, 'id': 652}, {'image_count': 1167, 'id': 653}, {'image_count': 15, 'id': 654}, {'image_count': 994, 'id': 655}, {'image_count': 28, 'id': 656}, {'image_count': 2, 'id': 657}, {'image_count': 338, 'id': 658}, {'image_count': 334, 'id': 659}, {'image_count': 15, 'id': 660}, {'image_count': 102, 'id': 661}, {'image_count': 1, 'id': 662}, {'image_count': 8, 'id': 663}, {'image_count': 1, 'id': 664}, {'image_count': 1, 'id': 665}, {'image_count': 28, 'id': 666}, {'image_count': 91, 'id': 667}, {'image_count': 260, 'id': 668}, {'image_count': 131, 'id': 669}, {'image_count': 128, 'id': 670}, {'image_count': 3, 'id': 671}, {'image_count': 10, 'id': 672}, {'image_count': 39, 'id': 673}, {'image_count': 2, 'id': 674}, {'image_count': 925, 'id': 675}, {'image_count': 354, 'id': 676}, {'image_count': 31, 'id': 677}, {'image_count': 10, 'id': 678}, {'image_count': 215, 'id': 679}, {'image_count': 71, 'id': 680}, {'image_count': 43, 'id': 681}, {'image_count': 28, 'id': 682}, {'image_count': 34, 'id': 683}, {'image_count': 16, 'id': 684}, {'image_count': 273, 'id': 685}, {'image_count': 2, 'id': 686}, {'image_count': 1388, 'id': 687}, {'image_count': 4, 'id': 688}, {'image_count': 107, 'id': 689}, {'image_count': 2, 'id': 690}, {'image_count': 1, 'id': 691}, {'image_count': 454, 'id': 692}, {'image_count': 9, 'id': 693}, {'image_count': 1901, 'id': 694}, {'image_count': 61, 'id': 695}, {'image_count': 91, 'id': 696}, {'image_count': 46, 'id': 697}, {'image_count': 1402, 'id': 698}, {'image_count': 74, 'id': 699}, {'image_count': 421, 'id': 700}, {'image_count': 226, 'id': 701}, {'image_count': 10, 'id': 702}, {'image_count': 3011, 'id': 703}, {'image_count': 261, 'id': 704}, {'image_count': 1671, 'id': 705}, {'image_count': 293, 'id': 706}, {'image_count': 62, 'id': 707}, {'image_count': 814, 'id': 708}, {'image_count': 407, 'id': 709}, {'image_count': 6, 'id': 710}, {'image_count': 16, 'id': 711}, {'image_count': 7, 'id': 712}, {'image_count': 1791, 'id': 713}, {'image_count': 2, 'id': 714}, {'image_count': 1915, 'id': 715}, {'image_count': 3327, 'id': 716}, {'image_count': 13, 'id': 717}, {'image_count': 16, 'id': 718}, {'image_count': 448, 'id': 719}, {'image_count': 12, 'id': 720}, {'image_count': 18, 'id': 721}, {'image_count': 4, 'id': 722}, {'image_count': 71, 'id': 723}, {'image_count': 189, 'id': 724}, {'image_count': 74, 'id': 725}, {'image_count': 103, 'id': 726}, {'image_count': 3, 'id': 727}, {'image_count': 110, 'id': 728}, {'image_count': 5, 'id': 729}, {'image_count': 9, 'id': 730}, {'image_count': 15, 'id': 731}, {'image_count': 25, 'id': 732}, {'image_count': 7, 'id': 733}, {'image_count': 647, 'id': 734}, {'image_count': 1510, 'id': 735}, {'image_count': 100, 'id': 736}, {'image_count': 47, 'id': 737}, {'image_count': 121, 'id': 738}, {'image_count': 2530, 'id': 739}, {'image_count': 73, 'id': 740}, {'image_count': 49, 'id': 741}, {'image_count': 23, 'id': 742}, {'image_count': 4, 'id': 743}, {'image_count': 62, 'id': 744}, {'image_count': 118, 'id': 745}, {'image_count': 99, 'id': 746}, {'image_count': 40, 'id': 747}, {'image_count': 1036, 'id': 748}, {'image_count': 105, 'id': 749}, {'image_count': 21, 'id': 750}, {'image_count': 229, 'id': 751}, {'image_count': 7, 'id': 752}, {'image_count': 72, 'id': 753}, {'image_count': 9, 'id': 754}, {'image_count': 10, 'id': 755}, {'image_count': 328, 'id': 756}, {'image_count': 468, 'id': 757}, {'image_count': 1, 'id': 758}, {'image_count': 2, 'id': 759}, {'image_count': 24, 'id': 760}, {'image_count': 11, 'id': 761}, {'image_count': 72, 'id': 762}, {'image_count': 17, 'id': 763}, {'image_count': 10, 'id': 764}, {'image_count': 17, 'id': 765}, {'image_count': 626, 'id': 766}, {'image_count': 47, 'id': 767}, {'image_count': 93, 'id': 768}, {'image_count': 1, 'id': 769}, {'image_count': 12, 'id': 770}, {'image_count': 228, 'id': 771}, {'image_count': 5, 'id': 772}, {'image_count': 76, 'id': 773}, {'image_count': 71, 'id': 774}, {'image_count': 30, 'id': 775}, {'image_count': 109, 'id': 776}, {'image_count': 14, 'id': 777}, {'image_count': 1, 'id': 778}, {'image_count': 8, 'id': 779}, {'image_count': 26, 'id': 780}, {'image_count': 339, 'id': 781}, {'image_count': 153, 'id': 782}, {'image_count': 2, 'id': 783}, {'image_count': 3, 'id': 784}, {'image_count': 8, 'id': 785}, {'image_count': 47, 'id': 786}, {'image_count': 8, 'id': 787}, {'image_count': 6, 'id': 788}, {'image_count': 116, 'id': 789}, {'image_count': 69, 'id': 790}, {'image_count': 13, 'id': 791}, {'image_count': 6, 'id': 792}, {'image_count': 54790, 'id': 793}, {'image_count': 79, 'id': 794}, {'image_count': 14, 'id': 795}, {'image_count': 7, 'id': 796}, {'image_count': 20, 'id': 797}, {'image_count': 114, 'id': 798}, {'image_count': 221, 'id': 799}, {'image_count': 502, 'id': 800}, {'image_count': 62, 'id': 801}, {'image_count': 87, 'id': 802}, {'image_count': 4, 'id': 803}, {'image_count': 1912, 'id': 804}, {'image_count': 7, 'id': 805}, {'image_count': 186, 'id': 806}, {'image_count': 18, 'id': 807}, {'image_count': 4, 'id': 808}, {'image_count': 3, 'id': 809}, {'image_count': 7, 'id': 810}, {'image_count': 1413, 'id': 811}, {'image_count': 7, 'id': 812}, {'image_count': 12, 'id': 813}, {'image_count': 248, 'id': 814}, {'image_count': 4, 'id': 815}, {'image_count': 2732, 'id': 816}, {'image_count': 529, 'id': 817}, {'image_count': 1932, 'id': 818}, {'image_count': 50, 'id': 819}, {'image_count': 3, 'id': 820}, {'image_count': 28, 'id': 821}, {'image_count': 10, 'id': 822}, {'image_count': 5, 'id': 823}, {'image_count': 5, 'id': 824}, {'image_count': 18, 'id': 825}, {'image_count': 14, 'id': 826}, {'image_count': 1890, 'id': 827}, {'image_count': 660, 'id': 828}, {'image_count': 8, 'id': 829}, {'image_count': 25, 'id': 830}, {'image_count': 10, 'id': 831}, {'image_count': 218, 'id': 832}, {'image_count': 36, 'id': 833}, {'image_count': 16, 'id': 834}, {'image_count': 808, 'id': 835}, {'image_count': 479, 'id': 836}, {'image_count': 3999, 'id': 837}, {'image_count': 307, 'id': 838}, {'image_count': 57, 'id': 839}, {'image_count': 28, 'id': 840}, {'image_count': 80, 'id': 841}, {'image_count': 11, 'id': 842}, {'image_count': 92, 'id': 843}, {'image_count': 20, 'id': 844}, {'image_count': 194, 'id': 845}, {'image_count': 23, 'id': 846}, {'image_count': 52, 'id': 847}, {'image_count': 673, 'id': 848}, {'image_count': 2, 'id': 849}, {'image_count': 2, 'id': 850}, {'image_count': 1, 'id': 851}, {'image_count': 2, 'id': 852}, {'image_count': 8, 'id': 853}, {'image_count': 80, 'id': 854}, {'image_count': 3, 'id': 855}, {'image_count': 3, 'id': 856}, {'image_count': 15, 'id': 857}, {'image_count': 2, 'id': 858}, {'image_count': 10, 'id': 859}, {'image_count': 386, 'id': 860}, {'image_count': 65, 'id': 861}, {'image_count': 3, 'id': 862}, {'image_count': 35, 'id': 863}, {'image_count': 5, 'id': 864}, {'image_count': 180, 'id': 865}, {'image_count': 99, 'id': 866}, {'image_count': 49, 'id': 867}, {'image_count': 28, 'id': 868}, {'image_count': 1, 'id': 869}, {'image_count': 52, 'id': 870}, {'image_count': 36, 'id': 871}, {'image_count': 70, 'id': 872}, {'image_count': 6, 'id': 873}, {'image_count': 29, 'id': 874}, {'image_count': 24, 'id': 875}, {'image_count': 1115, 'id': 876}, {'image_count': 61, 'id': 877}, {'image_count': 18, 'id': 878}, {'image_count': 18, 'id': 879}, {'image_count': 665, 'id': 880}, {'image_count': 2753, 'id': 881}, {'image_count': 29, 'id': 882}, {'image_count': 8, 'id': 883}, {'image_count': 14, 'id': 884}, {'image_count': 1622, 'id': 885}, {'image_count': 2, 'id': 886}, {'image_count': 3, 'id': 887}, {'image_count': 32, 'id': 888}, {'image_count': 55, 'id': 889}, {'image_count': 1, 'id': 890}, {'image_count': 10, 'id': 891}, {'image_count': 10, 'id': 892}, {'image_count': 47, 'id': 893}, {'image_count': 3, 'id': 894}, {'image_count': 29, 'id': 895}, {'image_count': 342, 'id': 896}, {'image_count': 25, 'id': 897}, {'image_count': 1469, 'id': 898}, {'image_count': 521, 'id': 899}, {'image_count': 347, 'id': 900}, {'image_count': 35, 'id': 901}, {'image_count': 7, 'id': 902}, {'image_count': 207, 'id': 903}, {'image_count': 108, 'id': 904}, {'image_count': 2, 'id': 905}, {'image_count': 34, 'id': 906}, {'image_count': 12, 'id': 907}, {'image_count': 10, 'id': 908}, {'image_count': 13, 'id': 909}, {'image_count': 361, 'id': 910}, {'image_count': 1023, 'id': 911}, {'image_count': 2074, 'id': 912}, {'image_count': 2, 'id': 913}, {'image_count': 5, 'id': 914}, {'image_count': 247, 'id': 915}, {'image_count': 1197, 'id': 916}, {'image_count': 4, 'id': 917}, {'image_count': 8, 'id': 918}, {'image_count': 158, 'id': 919}, {'image_count': 3, 'id': 920}, {'image_count': 752, 'id': 921}, {'image_count': 64, 'id': 922}, {'image_count': 840, 'id': 923}, {'image_count': 143, 'id': 924}, {'image_count': 1, 'id': 925}, {'image_count': 49, 'id': 926}, {'image_count': 126, 'id': 927}, {'image_count': 76, 'id': 928}, {'image_count': 11, 'id': 929}, {'image_count': 11, 'id': 930}, {'image_count': 4, 'id': 931}, {'image_count': 39, 'id': 932}, {'image_count': 11, 'id': 933}, {'image_count': 13, 'id': 934}, {'image_count': 91, 'id': 935}, {'image_count': 14, 'id': 936}, {'image_count': 5, 'id': 937}, {'image_count': 3, 'id': 938}, {'image_count': 10, 'id': 939}, {'image_count': 18, 'id': 940}, {'image_count': 9, 'id': 941}, {'image_count': 6, 'id': 942}, {'image_count': 1283, 'id': 943}, {'image_count': 2, 'id': 944}, {'image_count': 1, 'id': 945}, {'image_count': 19, 'id': 946}, {'image_count': 1942, 'id': 947}, {'image_count': 1916, 'id': 948}, {'image_count': 139, 'id': 949}, {'image_count': 43, 'id': 950}, {'image_count': 1969, 'id': 951}, {'image_count': 5, 'id': 952}, {'image_count': 134, 'id': 953}, {'image_count': 74, 'id': 954}, {'image_count': 381, 'id': 955}, {'image_count': 1, 'id': 956}, {'image_count': 381, 'id': 957}, {'image_count': 6, 'id': 958}, {'image_count': 1826, 'id': 959}, {'image_count': 28, 'id': 960}, {'image_count': 4082, 'id': 961}, {'image_count': 2943, 'id': 962}, {'image_count': 16, 'id': 963}, {'image_count': 2716, 'id': 964}, {'image_count': 1789, 'id': 965}, {'image_count': 401, 'id': 966}, {'image_count': 1968, 'id': 967}, {'image_count': 1167, 'id': 968}, {'image_count': 1, 'id': 969}, {'image_count': 56, 'id': 970}, {'image_count': 17, 'id': 971}, {'image_count': 1, 'id': 972}, {'image_count': 58, 'id': 973}, {'image_count': 9, 'id': 974}, {'image_count': 8, 'id': 975}, {'image_count': 1438, 'id': 976}, {'image_count': 31, 'id': 977}, {'image_count': 16, 'id': 978}, {'image_count': 491, 'id': 979}, {'image_count': 432, 'id': 980}, {'image_count': 1945, 'id': 981}, {'image_count': 3826, 'id': 982}, {'image_count': 5, 'id': 983}, {'image_count': 28, 'id': 984}, {'image_count': 7, 'id': 985}, {'image_count': 146, 'id': 986}, {'image_count': 1, 'id': 987}, {'image_count': 25, 'id': 988}, {'image_count': 22, 'id': 989}, {'image_count': 1, 'id': 990}, {'image_count': 10, 'id': 991}, {'image_count': 9, 'id': 992}, {'image_count': 308, 'id': 993}, {'image_count': 4, 'id': 994}, {'image_count': 1969, 'id': 995}, {'image_count': 45, 'id': 996}, {'image_count': 12, 'id': 997}, {'image_count': 1, 'id': 998}, {'image_count': 85, 'id': 999}, {'image_count': 3117, 'id': 1000}, {'image_count': 11, 'id': 1001}, {'image_count': 60, 'id': 1002}, {'image_count': 1, 'id': 1003}, {'image_count': 16, 'id': 1004}, {'image_count': 1, 'id': 1005}, {'image_count': 65, 'id': 1006}, {'image_count': 13, 'id': 1007}, {'image_count': 655, 'id': 1008}, {'image_count': 51, 'id': 1009}, {'image_count': 1, 'id': 1010}, {'image_count': 673, 'id': 1011}, {'image_count': 5, 'id': 1012}, {'image_count': 36, 'id': 1013}, {'image_count': 54, 'id': 1014}, {'image_count': 5, 'id': 1015}, {'image_count': 8, 'id': 1016}, {'image_count': 305, 'id': 1017}, {'image_count': 297, 'id': 1018}, {'image_count': 1508, 'id': 1019}, {'image_count': 223, 'id': 1020}, {'image_count': 1037, 'id': 1021}, {'image_count': 63, 'id': 1022}, {'image_count': 1881, 'id': 1023}, {'image_count': 507, 'id': 1024}, {'image_count': 333, 'id': 1025}, {'image_count': 1911, 'id': 1026}, {'image_count': 1765, 'id': 1027}, {'image_count': 1, 'id': 1028}, {'image_count': 5, 'id': 1029}, {'image_count': 1, 'id': 1030}, {'image_count': 9, 'id': 1031}, {'image_count': 2, 'id': 1032}, {'image_count': 151, 'id': 1033}, {'image_count': 82, 'id': 1034}, {'image_count': 1931, 'id': 1035}, {'image_count': 41, 'id': 1036}, {'image_count': 2877, 'id': 1037}, {'image_count': 24, 'id': 1038}, {'image_count': 22, 'id': 1039}, {'image_count': 35, 'id': 1040}, {'image_count': 69, 'id': 1041}, {'image_count': 962, 'id': 1042}, {'image_count': 588, 'id': 1043}, {'image_count': 21, 'id': 1044}, {'image_count': 825, 'id': 1045}, {'image_count': 52, 'id': 1046}, {'image_count': 5, 'id': 1047}, {'image_count': 5, 'id': 1048}, {'image_count': 5, 'id': 1049}, {'image_count': 1860, 'id': 1050}, {'image_count': 56, 'id': 1051}, {'image_count': 1582, 'id': 1052}, {'image_count': 7, 'id': 1053}, {'image_count': 2, 'id': 1054}, {'image_count': 1562, 'id': 1055}, {'image_count': 1885, 'id': 1056}, {'image_count': 1, 'id': 1057}, {'image_count': 5, 'id': 1058}, {'image_count': 137, 'id': 1059}, {'image_count': 1094, 'id': 1060}, {'image_count': 134, 'id': 1061}, {'image_count': 29, 'id': 1062}, {'image_count': 22, 'id': 1063}, {'image_count': 522, 'id': 1064}, {'image_count': 50, 'id': 1065}, {'image_count': 68, 'id': 1066}, {'image_count': 16, 'id': 1067}, {'image_count': 40, 'id': 1068}, {'image_count': 35, 'id': 1069}, {'image_count': 135, 'id': 1070}, {'image_count': 1866, 'id': 1071}, {'image_count': 772, 'id': 1072}, {'image_count': 50, 'id': 1073}, {'image_count': 1015, 'id': 1074}, {'image_count': 1, 'id': 1075}, {'image_count': 65, 'id': 1076}, {'image_count': 3978, 'id': 1077}, {'image_count': 1302, 'id': 1078}, {'image_count': 2925, 'id': 1079}, {'image_count': 2, 'id': 1080}, {'image_count': 29, 'id': 1081}, {'image_count': 36, 'id': 1082}, {'image_count': 138, 'id': 1083}, {'image_count': 4, 'id': 1084}, {'image_count': 67, 'id': 1085}, {'image_count': 26, 'id': 1086}, {'image_count': 25, 'id': 1087}, {'image_count': 33, 'id': 1088}, {'image_count': 37, 'id': 1089}, {'image_count': 50, 'id': 1090}, {'image_count': 270, 'id': 1091}, {'image_count': 12, 'id': 1092}, {'image_count': 316, 'id': 1093}, {'image_count': 41, 'id': 1094}, {'image_count': 299, 'id': 1095}, {'image_count': 105, 'id': 1096}, {'image_count': 2879, 'id': 1097}, {'image_count': 1021, 'id': 1098}, {'image_count': 1213, 'id': 1099}, {'image_count': 172, 'id': 1100}, {'image_count': 28, 'id': 1101}, {'image_count': 899, 'id': 1102}, {'image_count': 187, 'id': 1103}, {'image_count': 147, 'id': 1104}, {'image_count': 136, 'id': 1105}, {'image_count': 34, 'id': 1106}, {'image_count': 41, 'id': 1107}, {'image_count': 636, 'id': 1108}, {'image_count': 570, 'id': 1109}, {'image_count': 1149, 'id': 1110}, {'image_count': 61, 'id': 1111}, {'image_count': 3556, 'id': 1112}, {'image_count': 18, 'id': 1113}, {'image_count': 143, 'id': 1114}, {'image_count': 2951, 'id': 1115}, {'image_count': 7, 'id': 1116}, {'image_count': 943, 'id': 1117}, {'image_count': 6, 'id': 1118}, {'image_count': 1, 'id': 1119}, {'image_count': 11, 'id': 1120}, {'image_count': 101, 'id': 1121}, {'image_count': 1909, 'id': 1122}, {'image_count': 5302, 'id': 1123}, {'image_count': 1, 'id': 1124}, {'image_count': 44, 'id': 1125}, {'image_count': 3, 'id': 1126}, {'image_count': 44, 'id': 1127}, {'image_count': 31, 'id': 1128}, {'image_count': 7, 'id': 1129}, {'image_count': 20, 'id': 1130}, {'image_count': 11, 'id': 1131}, {'image_count': 13, 'id': 1132}, {'image_count': 3452, 'id': 1133}, {'image_count': 113, 'id': 1134}, {'image_count': 2, 'id': 1135}, {'image_count': 139, 'id': 1136}, {'image_count': 12, 'id': 1137}, {'image_count': 37, 'id': 1138}, {'image_count': 3213, 'id': 1139}, {'image_count': 47, 'id': 1140}, {'image_count': 1468, 'id': 1141}, {'image_count': 729, 'id': 1142}, {'image_count': 24, 'id': 1143}, {'image_count': 1, 'id': 1144}, {'image_count': 10, 'id': 1145}, {'image_count': 3, 'id': 1146}, {'image_count': 14, 'id': 1147}, {'image_count': 4, 'id': 1148}, {'image_count': 29, 'id': 1149}, {'image_count': 4, 'id': 1150}, {'image_count': 70, 'id': 1151}, {'image_count': 46, 'id': 1152}, {'image_count': 14, 'id': 1153}, {'image_count': 48, 'id': 1154}, {'image_count': 1855, 'id': 1155}, {'image_count': 113, 'id': 1156}, {'image_count': 1, 'id': 1157}, {'image_count': 1, 'id': 1158}, {'image_count': 10, 'id': 1159}, {'image_count': 54, 'id': 1160}, {'image_count': 1923, 'id': 1161}, {'image_count': 630, 'id': 1162}, {'image_count': 31, 'id': 1163}, {'image_count': 69, 'id': 1164}, {'image_count': 7, 'id': 1165}, {'image_count': 11, 'id': 1166}, {'image_count': 1, 'id': 1167}, {'image_count': 30, 'id': 1168}, {'image_count': 50, 'id': 1169}, {'image_count': 45, 'id': 1170}, {'image_count': 28, 'id': 1171}, {'image_count': 114, 'id': 1172}, {'image_count': 193, 'id': 1173}, {'image_count': 21, 'id': 1174}, {'image_count': 91, 'id': 1175}, {'image_count': 31, 'id': 1176}, {'image_count': 1469, 'id': 1177}, {'image_count': 1924, 'id': 1178}, {'image_count': 87, 'id': 1179}, {'image_count': 77, 'id': 1180}, {'image_count': 11, 'id': 1181}, {'image_count': 47, 'id': 1182}, {'image_count': 21, 'id': 1183}, {'image_count': 47, 'id': 1184}, {'image_count': 70, 'id': 1185}, {'image_count': 1838, 'id': 1186}, {'image_count': 19, 'id': 1187}, {'image_count': 531, 'id': 1188}, {'image_count': 11, 'id': 1189}, {'image_count': 2179, 'id': 1190}, {'image_count': 113, 'id': 1191}, {'image_count': 26, 'id': 1192}, {'image_count': 5, 'id': 1193}, {'image_count': 56, 'id': 1194}, {'image_count': 73, 'id': 1195}, {'image_count': 32, 'id': 1196}, {'image_count': 128, 'id': 1197}, {'image_count': 623, 'id': 1198}, {'image_count': 12, 'id': 1199}, {'image_count': 52, 'id': 1200}, {'image_count': 11, 'id': 1201}, {'image_count': 1687, 'id': 1202}, {'image_count': 81, 'id': 1203}]  # noqa
+# fmt: on

approach/ovod/APE/ape/data/datasets/objects365.py

@@ -0,0 +1,799 @@
+import os
+
+from detectron2.data.datasets.register_coco import register_coco_instances
+
+OBJECTS365_CATEGORIES_FIXNAME = [
+    {"id": 1, "name": "Person"},
+    {"id": 2, "name": "Sneakers"},
+    {"id": 3, "name": "Chair"},
+    {"id": 4, "name": "Other Shoes"},
+    {"id": 5, "name": "Hat"},
+    {"id": 6, "name": "Car"},
+    {"id": 7, "name": "Lamp"},
+    {"id": 8, "name": "Glasses"},
+    {"id": 9, "name": "Bottle"},
+    {"id": 10, "name": "Desk"},
+    {"id": 11, "name": "Cup"},
+    {"id": 12, "name": "Street Lights"},
+    {"id": 13, "name": "Cabinet/shelf"},
+    {"id": 14, "name": "Handbag/Satchel"},
+    {"id": 15, "name": "Bracelet"},
+    {"id": 16, "name": "Plate"},
+    {"id": 17, "name": "Picture/Frame"},
+    {"id": 18, "name": "Helmet"},
+    {"id": 19, "name": "Book"},
+    {"id": 20, "name": "Gloves"},
+    {"id": 21, "name": "Storage box"},
+    {"id": 22, "name": "Boat"},
+    {"id": 23, "name": "Leather Shoes"},
+    {"id": 24, "name": "Flower"},
+    {"id": 25, "name": "Bench"},
+    {"id": 26, "name": "Potted Plant"},
+    {"id": 27, "name": "Bowl/Basin"},
+    {"id": 28, "name": "Flag"},
+    {"id": 29, "name": "Pillow"},
+    {"id": 30, "name": "Boots"},
+    {"id": 31, "name": "Vase"},
+    {"id": 32, "name": "Microphone"},
+    {"id": 33, "name": "Necklace"},
+    {"id": 34, "name": "Ring"},
+    {"id": 35, "name": "SUV"},
+    {"id": 36, "name": "Wine Glass"},
+    {"id": 37, "name": "Belt"},
+    {"id": 38, "name": "Monitor/TV"},
+    {"id": 39, "name": "Backpack"},
+    {"id": 40, "name": "Umbrella"},
+    {"id": 41, "name": "Traffic Light"},
+    {"id": 42, "name": "Speaker"},
+    {"id": 43, "name": "Watch"},
+    {"id": 44, "name": "Tie"},
+    {"id": 45, "name": "Trash bin Can"},
+    {"id": 46, "name": "Slippers"},
+    {"id": 47, "name": "Bicycle"},
+    {"id": 48, "name": "Stool"},
+    {"id": 49, "name": "Barrel/bucket"},
+    {"id": 50, "name": "Van"},
+    {"id": 51, "name": "Couch"},
+    {"id": 52, "name": "Sandals"},
+    {"id": 53, "name": "Basket"},
+    {"id": 54, "name": "Drum"},
+    {"id": 55, "name": "Pen/Pencil"},
+    {"id": 56, "name": "Bus"},
+    {"id": 57, "name": "Wild Bird"},
+    {"id": 58, "name": "High Heels"},
+    {"id": 59, "name": "Motorcycle"},
+    {"id": 60, "name": "Guitar"},
+    {"id": 61, "name": "Carpet"},
+    {"id": 62, "name": "Cell Phone"},
+    {"id": 63, "name": "Bread"},
+    {"id": 64, "name": "Camera"},
+    {"id": 65, "name": "Canned"},
+    {"id": 66, "name": "Truck"},
+    {"id": 67, "name": "Traffic cone"},
+    {"id": 68, "name": "Cymbal"},
+    {"id": 69, "name": "Lifesaver"},
+    {"id": 70, "name": "Towel"},
+    {"id": 71, "name": "Stuffed Toy"},
+    {"id": 72, "name": "Candle"},
+    {"id": 73, "name": "Sailboat"},
+    {"id": 74, "name": "Laptop"},
+    {"id": 75, "name": "Awning"},
+    {"id": 76, "name": "Bed"},
+    {"id": 77, "name": "Faucet"},
+    {"id": 78, "name": "Tent"},
+    {"id": 79, "name": "Horse"},
+    {"id": 80, "name": "Mirror"},
+    {"id": 81, "name": "Power outlet"},
+    {"id": 82, "name": "Sink"},
+    {"id": 83, "name": "Apple"},
+    {"id": 84, "name": "Air Conditioner"},
+    {"id": 85, "name": "Knife"},
+    {"id": 86, "name": "Hockey Stick"},
+    {"id": 87, "name": "Paddle"},
+    {"id": 88, "name": "Pickup Truck"},
+    {"id": 89, "name": "Fork"},
+    {"id": 90, "name": "Traffic Sign"},
+    {"id": 91, "name": "Ballon"},
+    {"id": 92, "name": "Tripod"},
+    {"id": 93, "name": "Dog"},
+    {"id": 94, "name": "Spoon"},
+    {"id": 95, "name": "Clock"},
+    {"id": 96, "name": "Pot"},
+    {"id": 97, "name": "Cow"},
+    {"id": 98, "name": "Cake"},
+    {"id": 99, "name": "Dining Table"},
+    {"id": 100, "name": "Sheep"},
+    {"id": 101, "name": "Hanger"},
+    {"id": 102, "name": "Blackboard/Whiteboard"},
+    {"id": 103, "name": "Napkin"},
+    {"id": 104, "name": "Other Fish"},
+    {"id": 105, "name": "Orange/Tangerine"},
+    {"id": 106, "name": "Toiletry"},
+    {"id": 107, "name": "Keyboard"},
+    {"id": 108, "name": "Tomato"},
+    {"id": 109, "name": "Lantern"},
+    {"id": 110, "name": "Machinery Vehicle"},
+    {"id": 111, "name": "Fan"},
+    {"id": 112, "name": "Green Vegetables"},
+    {"id": 113, "name": "Banana"},
+    {"id": 114, "name": "Baseball Glove"},
+    {"id": 115, "name": "Airplane"},
+    {"id": 116, "name": "Mouse"},
+    {"id": 117, "name": "Train"},
+    {"id": 118, "name": "Pumpkin"},
+    {"id": 119, "name": "Soccer"},
+    {"id": 120, "name": "Skiboard"},
+    {"id": 121, "name": "Luggage"},
+    {"id": 122, "name": "Nightstand"},
+    {"id": 123, "name": "Teapot"},
+    {"id": 124, "name": "Telephone"},
+    {"id": 125, "name": "Trolley"},
+    {"id": 126, "name": "Head Phone"},
+    {"id": 127, "name": "Sports Car"},
+    {"id": 128, "name": "Stop Sign"},
+    {"id": 129, "name": "Dessert"},
+    {"id": 130, "name": "Scooter"},
+    {"id": 131, "name": "Stroller"},
+    {"id": 132, "name": "Crane"},
+    {"id": 133, "name": "Remote"},
+    {"id": 134, "name": "Refrigerator"},
+    {"id": 135, "name": "Oven"},
+    {"id": 136, "name": "Lemon"},
+    {"id": 137, "name": "Duck"},
+    {"id": 138, "name": "Baseball Bat"},
+    {"id": 139, "name": "Surveillance Camera"},
+    {"id": 140, "name": "Cat"},
+    {"id": 141, "name": "Jug"},
+    {"id": 142, "name": "Broccoli"},
+    {"id": 143, "name": "Piano"},
+    {"id": 144, "name": "Pizza"},
+    {"id": 145, "name": "Elephant"},
+    {"id": 146, "name": "Skateboard"},
+    {"id": 147, "name": "Surfboard"},
+    {"id": 148, "name": "Gun"},
+    {"id": 149, "name": "Skating and Skiing shoes"},
+    {"id": 150, "name": "Gas stove"},
+    {"id": 151, "name": "Donut"},
+    {"id": 152, "name": "Bow Tie"},
+    {"id": 153, "name": "Carrot"},
+    {"id": 154, "name": "Toilet"},
+    {"id": 155, "name": "Kite"},
+    {"id": 156, "name": "Strawberry"},
+    {"id": 157, "name": "Other Balls"},
+    {"id": 158, "name": "Shovel"},
+    {"id": 159, "name": "Pepper"},
+    {"id": 160, "name": "Computer Box"},
+    {"id": 161, "name": "Toilet Paper"},
+    {"id": 162, "name": "Cleaning Products"},
+    {"id": 163, "name": "Chopsticks"},
+    {"id": 164, "name": "Microwave"},
+    {"id": 165, "name": "Pigeon"},
+    {"id": 166, "name": "Baseball"},
+    {"id": 167, "name": "Cutting/chopping Board"},
+    {"id": 168, "name": "Coffee Table"},
+    {"id": 169, "name": "Side Table"},
+    {"id": 170, "name": "Scissors"},
+    {"id": 171, "name": "Marker"},
+    {"id": 172, "name": "Pie"},
+    {"id": 173, "name": "Ladder"},
+    {"id": 174, "name": "Snowboard"},
+    {"id": 175, "name": "Cookies"},
+    {"id": 176, "name": "Radiator"},
+    {"id": 177, "name": "Fire Hydrant"},
+    {"id": 178, "name": "Basketball"},
+    {"id": 179, "name": "Zebra"},
+    {"id": 180, "name": "Grape"},
+    {"id": 181, "name": "Giraffe"},
+    {"id": 182, "name": "Potato"},
+    {"id": 183, "name": "Sausage"},
+    {"id": 184, "name": "Tricycle"},
+    {"id": 185, "name": "Violin"},
+    {"id": 186, "name": "Egg"},
+    {"id": 187, "name": "Fire Extinguisher"},
+    {"id": 188, "name": "Candy"},
+    {"id": 189, "name": "Fire Truck"},
+    {"id": 190, "name": "Billards"},
+    {"id": 191, "name": "Converter"},
+    {"id": 192, "name": "Bathtub"},
+    {"id": 193, "name": "Wheelchair"},
+    {"id": 194, "name": "Golf Club"},
+    {"id": 195, "name": "Briefcase"},
+    {"id": 196, "name": "Cucumber"},
+    {"id": 197, "name": "Cigar/Cigarette "},
+    {"id": 198, "name": "Paint Brush"},
+    {"id": 199, "name": "Pear"},
+    {"id": 200, "name": "Heavy Truck"},
+    {"id": 201, "name": "Hamburger"},
+    {"id": 202, "name": "Extractor"},
+    {"id": 203, "name": "Extension Cord"},
+    {"id": 204, "name": "Tong"},
+    {"id": 205, "name": "Tennis Racket"},
+    {"id": 206, "name": "Folder"},
+    {"id": 207, "name": "American Football"},
+    {"id": 208, "name": "earphone"},
+    {"id": 209, "name": "Mask"},
+    {"id": 210, "name": "Kettle"},
+    {"id": 211, "name": "Tennis"},
+    {"id": 212, "name": "Ship"},
+    {"id": 213, "name": "Swing"},
+    {"id": 214, "name": "Coffee Machine"},
+    {"id": 215, "name": "Slide"},
+    {"id": 216, "name": "Carriage"},
+    {"id": 217, "name": "Onion"},
+    {"id": 218, "name": "Green beans"},
+    {"id": 219, "name": "Projector"},
+    {"id": 220, "name": "Frisbee"},
+    {"id": 221, "name": "Washing Machine/Drying Machine"},
+    {"id": 222, "name": "Chicken"},
+    {"id": 223, "name": "Printer"},
+    {"id": 224, "name": "Watermelon"},
+    {"id": 225, "name": "Saxophone"},
+    {"id": 226, "name": "Tissue"},
+    {"id": 227, "name": "Toothbrush"},
+    {"id": 228, "name": "Ice cream"},
+    {"id": 229, "name": "Hot air balloon"},
+    {"id": 230, "name": "Cello"},
+    {"id": 231, "name": "French Fries"},
+    {"id": 232, "name": "Scale"},
+    {"id": 233, "name": "Trophy"},
+    {"id": 234, "name": "Cabbage"},
+    {"id": 235, "name": "Hot dog"},
+    {"id": 236, "name": "Blender"},
+    {"id": 237, "name": "Peach"},
+    {"id": 238, "name": "Rice"},
+    {"id": 239, "name": "Wallet/Purse"},
+    {"id": 240, "name": "Volleyball"},
+    {"id": 241, "name": "Deer"},
+    {"id": 242, "name": "Goose"},
+    {"id": 243, "name": "Tape"},
+    {"id": 244, "name": "Tablet"},
+    {"id": 245, "name": "Cosmetics"},
+    {"id": 246, "name": "Trumpet"},
+    {"id": 247, "name": "Pineapple"},
+    {"id": 248, "name": "Golf Ball"},
+    {"id": 249, "name": "Ambulance"},
+    {"id": 250, "name": "Parking meter"},
+    {"id": 251, "name": "Mango"},
+    {"id": 252, "name": "Key"},
+    {"id": 253, "name": "Hurdle"},
+    {"id": 254, "name": "Fishing Rod"},
+    {"id": 255, "name": "Medal"},
+    {"id": 256, "name": "Flute"},
+    {"id": 257, "name": "Brush"},
+    {"id": 258, "name": "Penguin"},
+    {"id": 259, "name": "Megaphone"},
+    {"id": 260, "name": "Corn"},
+    {"id": 261, "name": "Lettuce"},
+    {"id": 262, "name": "Garlic"},
+    {"id": 263, "name": "Swan"},
+    {"id": 264, "name": "Helicopter"},
+    {"id": 265, "name": "Green Onion"},
+    {"id": 266, "name": "Sandwich"},
+    {"id": 267, "name": "Nuts"},
+    {"id": 268, "name": "Speed Limit Sign"},
+    {"id": 269, "name": "Induction Cooker"},
+    {"id": 270, "name": "Broom"},
+    {"id": 271, "name": "Trombone"},
+    {"id": 272, "name": "Plum"},
+    {"id": 273, "name": "Rickshaw"},
+    {"id": 274, "name": "Goldfish"},
+    {"id": 275, "name": "Kiwi fruit"},
+    {"id": 276, "name": "Router/modem"},
+    {"id": 277, "name": "Poker Card"},
+    {"id": 278, "name": "Toaster"},
+    {"id": 279, "name": "Shrimp"},
+    {"id": 280, "name": "Sushi"},
+    {"id": 281, "name": "Cheese"},
+    {"id": 282, "name": "Notepaper"},
+    {"id": 283, "name": "Cherry"},
+    {"id": 284, "name": "Pliers"},
+    {"id": 285, "name": "CD"},
+    {"id": 286, "name": "Pasta"},
+    {"id": 287, "name": "Hammer"},
+    {"id": 288, "name": "Cue"},
+    {"id": 289, "name": "Avocado"},
+    {"id": 290, "name": "Hami melon"},
+    {"id": 291, "name": "Flask"},
+    {"id": 292, "name": "Mushroom"},
+    {"id": 293, "name": "Screwdriver"},
+    {"id": 294, "name": "Soap"},
+    {"id": 295, "name": "Recorder"},
+    {"id": 296, "name": "Bear"},
+    {"id": 297, "name": "Eggplant"},
+    {"id": 298, "name": "Board Eraser"},
+    {"id": 299, "name": "Coconut"},
+    {"id": 300, "name": "Tape Measure/ Ruler"},
+    {"id": 301, "name": "Pig"},
+    {"id": 302, "name": "Showerhead"},
+    {"id": 303, "name": "Globe"},
+    {"id": 304, "name": "Chips"},
+    {"id": 305, "name": "Steak"},
+    {"id": 306, "name": "Crosswalk Sign"},
+    {"id": 307, "name": "Stapler"},
+    {"id": 308, "name": "Camel"},
+    {"id": 309, "name": "Formula 1 "},
+    {"id": 310, "name": "Pomegranate"},
+    {"id": 311, "name": "Dishwasher"},
+    {"id": 312, "name": "Crab"},
+    {"id": 313, "name": "Hoverboard"},
+    {"id": 314, "name": "Meatball"},
+    {"id": 315, "name": "Rice Cooker"},
+    {"id": 316, "name": "Tuba"},
+    {"id": 317, "name": "Calculator"},
+    {"id": 318, "name": "Papaya"},
+    {"id": 319, "name": "Antelope"},
+    {"id": 320, "name": "Parrot"},
+    {"id": 321, "name": "Seal"},
+    {"id": 322, "name": "Butterfly"},
+    {"id": 323, "name": "Dumbbell"},
+    {"id": 324, "name": "Donkey"},
+    {"id": 325, "name": "Lion"},
+    {"id": 326, "name": "Urinal"},
+    {"id": 327, "name": "Dolphin"},
+    {"id": 328, "name": "Electric Drill"},
+    {"id": 329, "name": "Hair Dryer"},
+    {"id": 330, "name": "Egg tart"},
+    {"id": 331, "name": "Jellyfish"},
+    {"id": 332, "name": "Treadmill"},
+    {"id": 333, "name": "Lighter"},
+    {"id": 334, "name": "Grapefruit"},
+    {"id": 335, "name": "Game board"},
+    {"id": 336, "name": "Mop"},
+    {"id": 337, "name": "Radish"},
+    {"id": 338, "name": "Baozi"},
+    {"id": 339, "name": "Target"},
+    {"id": 340, "name": "French"},
+    {"id": 341, "name": "Spring Rolls"},
+    {"id": 342, "name": "Monkey"},
+    {"id": 343, "name": "Rabbit"},
+    {"id": 344, "name": "Pencil Case"},
+    {"id": 345, "name": "Yak"},
+    {"id": 346, "name": "Red Cabbage"},
+    {"id": 347, "name": "Binoculars"},
+    {"id": 348, "name": "Asparagus"},
+    {"id": 349, "name": "Barbell"},
+    {"id": 350, "name": "Scallop"},
+    {"id": 351, "name": "Noddles"},
+    {"id": 352, "name": "Comb"},
+    {"id": 353, "name": "Dumpling"},
+    {"id": 354, "name": "Oyster"},
+    {"id": 355, "name": "Table Tennis paddle"},
+    {"id": 356, "name": "Cosmetics Brush/Eyeliner Pencil"},
+    {"id": 357, "name": "Chainsaw"},
+    {"id": 358, "name": "Eraser"},
+    {"id": 359, "name": "Lobster"},
+    {"id": 360, "name": "Durian"},
+    {"id": 361, "name": "Okra"},
+    {"id": 362, "name": "Lipstick"},
+    {"id": 363, "name": "Cosmetics Mirror"},
+    {"id": 364, "name": "Curling"},
+    {"id": 365, "name": "Table Tennis "},
+]
+
+OBJECTS365_CATEGORIES = [
+    {"id": 1, "name": "Person"},
+    {"id": 2, "name": "Sneakers"},
+    {"id": 3, "name": "Chair"},
+    {"id": 4, "name": "Other Shoes"},
+    {"id": 5, "name": "Hat"},
+    {"id": 6, "name": "Car"},
+    {"id": 7, "name": "Lamp"},
+    {"id": 8, "name": "Glasses"},
+    {"id": 9, "name": "Bottle"},
+    {"id": 10, "name": "Desk"},
+    {"id": 11, "name": "Cup"},
+    {"id": 12, "name": "Street Lights"},
+    {"id": 13, "name": "Cabinet/shelf"},
+    {"id": 14, "name": "Handbag/Satchel"},
+    {"id": 15, "name": "Bracelet"},
+    {"id": 16, "name": "Plate"},
+    {"id": 17, "name": "Picture/Frame"},
+    {"id": 18, "name": "Helmet"},
+    {"id": 19, "name": "Book"},
+    {"id": 20, "name": "Gloves"},
+    {"id": 21, "name": "Storage box"},
+    {"id": 22, "name": "Boat"},
+    {"id": 23, "name": "Leather Shoes"},
+    {"id": 24, "name": "Flower"},
+    {"id": 25, "name": "Bench"},
+    {"id": 26, "name": "Potted Plant"},
+    {"id": 27, "name": "Bowl/Basin"},
+    {"id": 28, "name": "Flag"},
+    {"id": 29, "name": "Pillow"},
+    {"id": 30, "name": "Boots"},
+    {"id": 31, "name": "Vase"},
+    {"id": 32, "name": "Microphone"},
+    {"id": 33, "name": "Necklace"},
+    {"id": 34, "name": "Ring"},
+    {"id": 35, "name": "SUV"},
+    {"id": 36, "name": "Wine Glass"},
+    {"id": 37, "name": "Belt"},
+    {"id": 38, "name": "Moniter/TV"},
+    {"id": 39, "name": "Backpack"},
+    {"id": 40, "name": "Umbrella"},
+    {"id": 41, "name": "Traffic Light"},
+    {"id": 42, "name": "Speaker"},
+    {"id": 43, "name": "Watch"},
+    {"id": 44, "name": "Tie"},
+    {"id": 45, "name": "Trash bin Can"},
+    {"id": 46, "name": "Slippers"},
+    {"id": 47, "name": "Bicycle"},
+    {"id": 48, "name": "Stool"},
+    {"id": 49, "name": "Barrel/bucket"},
+    {"id": 50, "name": "Van"},
+    {"id": 51, "name": "Couch"},
+    {"id": 52, "name": "Sandals"},
+    {"id": 53, "name": "Bakset"},
+    {"id": 54, "name": "Drum"},
+    {"id": 55, "name": "Pen/Pencil"},
+    {"id": 56, "name": "Bus"},
+    {"id": 57, "name": "Wild Bird"},
+    {"id": 58, "name": "High Heels"},
+    {"id": 59, "name": "Motorcycle"},
+    {"id": 60, "name": "Guitar"},
+    {"id": 61, "name": "Carpet"},
+    {"id": 62, "name": "Cell Phone"},
+    {"id": 63, "name": "Bread"},
+    {"id": 64, "name": "Camera"},
+    {"id": 65, "name": "Canned"},
+    {"id": 66, "name": "Truck"},
+    {"id": 67, "name": "Traffic cone"},
+    {"id": 68, "name": "Cymbal"},
+    {"id": 69, "name": "Lifesaver"},
+    {"id": 70, "name": "Towel"},
+    {"id": 71, "name": "Stuffed Toy"},
+    {"id": 72, "name": "Candle"},
+    {"id": 73, "name": "Sailboat"},
+    {"id": 74, "name": "Laptop"},
+    {"id": 75, "name": "Awning"},
+    {"id": 76, "name": "Bed"},
+    {"id": 77, "name": "Faucet"},
+    {"id": 78, "name": "Tent"},
+    {"id": 79, "name": "Horse"},
+    {"id": 80, "name": "Mirror"},
+    {"id": 81, "name": "Power outlet"},
+    {"id": 82, "name": "Sink"},
+    {"id": 83, "name": "Apple"},
+    {"id": 84, "name": "Air Conditioner"},
+    {"id": 85, "name": "Knife"},
+    {"id": 86, "name": "Hockey Stick"},
+    {"id": 87, "name": "Paddle"},
+    {"id": 88, "name": "Pickup Truck"},
+    {"id": 89, "name": "Fork"},
+    {"id": 90, "name": "Traffic Sign"},
+    {"id": 91, "name": "Ballon"},
+    {"id": 92, "name": "Tripod"},
+    {"id": 93, "name": "Dog"},
+    {"id": 94, "name": "Spoon"},
+    {"id": 95, "name": "Clock"},
+    {"id": 96, "name": "Pot"},
+    {"id": 97, "name": "Cow"},
+    {"id": 98, "name": "Cake"},
+    {"id": 99, "name": "Dinning Table"},
+    {"id": 100, "name": "Sheep"},
+    {"id": 101, "name": "Hanger"},
+    {"id": 102, "name": "Blackboard/Whiteboard"},
+    {"id": 103, "name": "Napkin"},
+    {"id": 104, "name": "Other Fish"},
+    {"id": 105, "name": "Orange/Tangerine"},
+    {"id": 106, "name": "Toiletry"},
+    {"id": 107, "name": "Keyboard"},
+    {"id": 108, "name": "Tomato"},
+    {"id": 109, "name": "Lantern"},
+    {"id": 110, "name": "Machinery Vehicle"},
+    {"id": 111, "name": "Fan"},
+    {"id": 112, "name": "Green Vegetables"},
+    {"id": 113, "name": "Banana"},
+    {"id": 114, "name": "Baseball Glove"},
+    {"id": 115, "name": "Airplane"},
+    {"id": 116, "name": "Mouse"},
+    {"id": 117, "name": "Train"},
+    {"id": 118, "name": "Pumpkin"},
+    {"id": 119, "name": "Soccer"},
+    {"id": 120, "name": "Skiboard"},
+    {"id": 121, "name": "Luggage"},
+    {"id": 122, "name": "Nightstand"},
+    {"id": 123, "name": "Tea pot"},
+    {"id": 124, "name": "Telephone"},
+    {"id": 125, "name": "Trolley"},
+    {"id": 126, "name": "Head Phone"},
+    {"id": 127, "name": "Sports Car"},
+    {"id": 128, "name": "Stop Sign"},
+    {"id": 129, "name": "Dessert"},
+    {"id": 130, "name": "Scooter"},
+    {"id": 131, "name": "Stroller"},
+    {"id": 132, "name": "Crane"},
+    {"id": 133, "name": "Remote"},
+    {"id": 134, "name": "Refrigerator"},
+    {"id": 135, "name": "Oven"},
+    {"id": 136, "name": "Lemon"},
+    {"id": 137, "name": "Duck"},
+    {"id": 138, "name": "Baseball Bat"},
+    {"id": 139, "name": "Surveillance Camera"},
+    {"id": 140, "name": "Cat"},
+    {"id": 141, "name": "Jug"},
+    {"id": 142, "name": "Broccoli"},
+    {"id": 143, "name": "Piano"},
+    {"id": 144, "name": "Pizza"},
+    {"id": 145, "name": "Elephant"},
+    {"id": 146, "name": "Skateboard"},
+    {"id": 147, "name": "Surfboard"},
+    {"id": 148, "name": "Gun"},
+    {"id": 149, "name": "Skating and Skiing shoes"},
+    {"id": 150, "name": "Gas stove"},
+    {"id": 151, "name": "Donut"},
+    {"id": 152, "name": "Bow Tie"},
+    {"id": 153, "name": "Carrot"},
+    {"id": 154, "name": "Toilet"},
+    {"id": 155, "name": "Kite"},
+    {"id": 156, "name": "Strawberry"},
+    {"id": 157, "name": "Other Balls"},
+    {"id": 158, "name": "Shovel"},
+    {"id": 159, "name": "Pepper"},
+    {"id": 160, "name": "Computer Box"},
+    {"id": 161, "name": "Toilet Paper"},
+    {"id": 162, "name": "Cleaning Products"},
+    {"id": 163, "name": "Chopsticks"},
+    {"id": 164, "name": "Microwave"},
+    {"id": 165, "name": "Pigeon"},
+    {"id": 166, "name": "Baseball"},
+    {"id": 167, "name": "Cutting/chopping Board"},
+    {"id": 168, "name": "Coffee Table"},
+    {"id": 169, "name": "Side Table"},
+    {"id": 170, "name": "Scissors"},
+    {"id": 171, "name": "Marker"},
+    {"id": 172, "name": "Pie"},
+    {"id": 173, "name": "Ladder"},
+    {"id": 174, "name": "Snowboard"},
+    {"id": 175, "name": "Cookies"},
+    {"id": 176, "name": "Radiator"},
+    {"id": 177, "name": "Fire Hydrant"},
+    {"id": 178, "name": "Basketball"},
+    {"id": 179, "name": "Zebra"},
+    {"id": 180, "name": "Grape"},
+    {"id": 181, "name": "Giraffe"},
+    {"id": 182, "name": "Potato"},
+    {"id": 183, "name": "Sausage"},
+    {"id": 184, "name": "Tricycle"},
+    {"id": 185, "name": "Violin"},
+    {"id": 186, "name": "Egg"},
+    {"id": 187, "name": "Fire Extinguisher"},
+    {"id": 188, "name": "Candy"},
+    {"id": 189, "name": "Fire Truck"},
+    {"id": 190, "name": "Billards"},
+    {"id": 191, "name": "Converter"},
+    {"id": 192, "name": "Bathtub"},
+    {"id": 193, "name": "Wheelchair"},
+    {"id": 194, "name": "Golf Club"},
+    {"id": 195, "name": "Briefcase"},
+    {"id": 196, "name": "Cucumber"},
+    {"id": 197, "name": "Cigar/Cigarette "},
+    {"id": 198, "name": "Paint Brush"},
+    {"id": 199, "name": "Pear"},
+    {"id": 200, "name": "Heavy Truck"},
+    {"id": 201, "name": "Hamburger"},
+    {"id": 202, "name": "Extractor"},
+    {"id": 203, "name": "Extention Cord"},
+    {"id": 204, "name": "Tong"},
+    {"id": 205, "name": "Tennis Racket"},
+    {"id": 206, "name": "Folder"},
+    {"id": 207, "name": "American Football"},
+    {"id": 208, "name": "earphone"},
+    {"id": 209, "name": "Mask"},
+    {"id": 210, "name": "Kettle"},
+    {"id": 211, "name": "Tennis"},
+    {"id": 212, "name": "Ship"},
+    {"id": 213, "name": "Swing"},
+    {"id": 214, "name": "Coffee Machine"},
+    {"id": 215, "name": "Slide"},
+    {"id": 216, "name": "Carriage"},
+    {"id": 217, "name": "Onion"},
+    {"id": 218, "name": "Green beans"},
+    {"id": 219, "name": "Projector"},
+    {"id": 220, "name": "Frisbee"},
+    {"id": 221, "name": "Washing Machine/Drying Machine"},
+    {"id": 222, "name": "Chicken"},
+    {"id": 223, "name": "Printer"},
+    {"id": 224, "name": "Watermelon"},
+    {"id": 225, "name": "Saxophone"},
+    {"id": 226, "name": "Tissue"},
+    {"id": 227, "name": "Toothbrush"},
+    {"id": 228, "name": "Ice cream"},
+    {"id": 229, "name": "Hotair ballon"},
+    {"id": 230, "name": "Cello"},
+    {"id": 231, "name": "French Fries"},
+    {"id": 232, "name": "Scale"},
+    {"id": 233, "name": "Trophy"},
+    {"id": 234, "name": "Cabbage"},
+    {"id": 235, "name": "Hot dog"},
+    {"id": 236, "name": "Blender"},
+    {"id": 237, "name": "Peach"},
+    {"id": 238, "name": "Rice"},
+    {"id": 239, "name": "Wallet/Purse"},
+    {"id": 240, "name": "Volleyball"},
+    {"id": 241, "name": "Deer"},
+    {"id": 242, "name": "Goose"},
+    {"id": 243, "name": "Tape"},
+    {"id": 244, "name": "Tablet"},
+    {"id": 245, "name": "Cosmetics"},
+    {"id": 246, "name": "Trumpet"},
+    {"id": 247, "name": "Pineapple"},
+    {"id": 248, "name": "Golf Ball"},
+    {"id": 249, "name": "Ambulance"},
+    {"id": 250, "name": "Parking meter"},
+    {"id": 251, "name": "Mango"},
+    {"id": 252, "name": "Key"},
+    {"id": 253, "name": "Hurdle"},
+    {"id": 254, "name": "Fishing Rod"},
+    {"id": 255, "name": "Medal"},
+    {"id": 256, "name": "Flute"},
+    {"id": 257, "name": "Brush"},
+    {"id": 258, "name": "Penguin"},
+    {"id": 259, "name": "Megaphone"},
+    {"id": 260, "name": "Corn"},
+    {"id": 261, "name": "Lettuce"},
+    {"id": 262, "name": "Garlic"},
+    {"id": 263, "name": "Swan"},
+    {"id": 264, "name": "Helicopter"},
+    {"id": 265, "name": "Green Onion"},
+    {"id": 266, "name": "Sandwich"},
+    {"id": 267, "name": "Nuts"},
+    {"id": 268, "name": "Speed Limit Sign"},
+    {"id": 269, "name": "Induction Cooker"},
+    {"id": 270, "name": "Broom"},
+    {"id": 271, "name": "Trombone"},
+    {"id": 272, "name": "Plum"},
+    {"id": 273, "name": "Rickshaw"},
+    {"id": 274, "name": "Goldfish"},
+    {"id": 275, "name": "Kiwi fruit"},
+    {"id": 276, "name": "Router/modem"},
+    {"id": 277, "name": "Poker Card"},
+    {"id": 278, "name": "Toaster"},
+    {"id": 279, "name": "Shrimp"},
+    {"id": 280, "name": "Sushi"},
+    {"id": 281, "name": "Cheese"},
+    {"id": 282, "name": "Notepaper"},
+    {"id": 283, "name": "Cherry"},
+    {"id": 284, "name": "Pliers"},
+    {"id": 285, "name": "CD"},
+    {"id": 286, "name": "Pasta"},
+    {"id": 287, "name": "Hammer"},
+    {"id": 288, "name": "Cue"},
+    {"id": 289, "name": "Avocado"},
+    {"id": 290, "name": "Hamimelon"},
+    {"id": 291, "name": "Flask"},
+    {"id": 292, "name": "Mushroon"},
+    {"id": 293, "name": "Screwdriver"},
+    {"id": 294, "name": "Soap"},
+    {"id": 295, "name": "Recorder"},
+    {"id": 296, "name": "Bear"},
+    {"id": 297, "name": "Eggplant"},
+    {"id": 298, "name": "Board Eraser"},
+    {"id": 299, "name": "Coconut"},
+    {"id": 300, "name": "Tape Measur/ Ruler"},
+    {"id": 301, "name": "Pig"},
+    {"id": 302, "name": "Showerhead"},
+    {"id": 303, "name": "Globe"},
+    {"id": 304, "name": "Chips"},
+    {"id": 305, "name": "Steak"},
+    {"id": 306, "name": "Crosswalk Sign"},
+    {"id": 307, "name": "Stapler"},
+    {"id": 308, "name": "Campel"},
+    {"id": 309, "name": "Formula 1 "},
+    {"id": 310, "name": "Pomegranate"},
+    {"id": 311, "name": "Dishwasher"},
+    {"id": 312, "name": "Crab"},
+    {"id": 313, "name": "Hoverboard"},
+    {"id": 314, "name": "Meat ball"},
+    {"id": 315, "name": "Rice Cooker"},
+    {"id": 316, "name": "Tuba"},
+    {"id": 317, "name": "Calculator"},
+    {"id": 318, "name": "Papaya"},
+    {"id": 319, "name": "Antelope"},
+    {"id": 320, "name": "Parrot"},
+    {"id": 321, "name": "Seal"},
+    {"id": 322, "name": "Buttefly"},
+    {"id": 323, "name": "Dumbbell"},
+    {"id": 324, "name": "Donkey"},
+    {"id": 325, "name": "Lion"},
+    {"id": 326, "name": "Urinal"},
+    {"id": 327, "name": "Dolphin"},
+    {"id": 328, "name": "Electric Drill"},
+    {"id": 329, "name": "Hair Dryer"},
+    {"id": 330, "name": "Egg tart"},
+    {"id": 331, "name": "Jellyfish"},
+    {"id": 332, "name": "Treadmill"},
+    {"id": 333, "name": "Lighter"},
+    {"id": 334, "name": "Grapefruit"},
+    {"id": 335, "name": "Game board"},
+    {"id": 336, "name": "Mop"},
+    {"id": 337, "name": "Radish"},
+    {"id": 338, "name": "Baozi"},
+    {"id": 339, "name": "Target"},
+    {"id": 340, "name": "French"},
+    {"id": 341, "name": "Spring Rolls"},
+    {"id": 342, "name": "Monkey"},
+    {"id": 343, "name": "Rabbit"},
+    {"id": 344, "name": "Pencil Case"},
+    {"id": 345, "name": "Yak"},
+    {"id": 346, "name": "Red Cabbage"},
+    {"id": 347, "name": "Binoculars"},
+    {"id": 348, "name": "Asparagus"},
+    {"id": 349, "name": "Barbell"},
+    {"id": 350, "name": "Scallop"},
+    {"id": 351, "name": "Noddles"},
+    {"id": 352, "name": "Comb"},
+    {"id": 353, "name": "Dumpling"},
+    {"id": 354, "name": "Oyster"},
+    {"id": 355, "name": "Table Teniis paddle"},
+    {"id": 356, "name": "Cosmetics Brush/Eyeliner Pencil"},
+    {"id": 357, "name": "Chainsaw"},
+    {"id": 358, "name": "Eraser"},
+    {"id": 359, "name": "Lobster"},
+    {"id": 360, "name": "Durian"},
+    {"id": 361, "name": "Okra"},
+    {"id": 362, "name": "Lipstick"},
+    {"id": 363, "name": "Cosmetics Mirror"},
+    {"id": 364, "name": "Curling"},
+    {"id": 365, "name": "Table Tennis "},
+]
+
+
+def _get_builtin_metadata(key):
+    # return {}
+    if "fixname" in key:
+        id_to_name = {x["id"]: x["name"] for x in OBJECTS365_CATEGORIES_FIXNAME}
+        thing_dataset_id_to_contiguous_id = {
+            i + 1: i for i in range(len(OBJECTS365_CATEGORIES_FIXNAME))
+        }
+    else:
+        id_to_name = {x["id"]: x["name"] for x in OBJECTS365_CATEGORIES}
+        thing_dataset_id_to_contiguous_id = {i + 1: i for i in range(len(OBJECTS365_CATEGORIES))}
+    thing_classes = [id_to_name[k] for k in sorted(id_to_name)]
+    return {
+        "thing_dataset_id_to_contiguous_id": thing_dataset_id_to_contiguous_id,
+        "thing_classes": thing_classes,
+    }
+
+
+_PREDEFINED_SPLITS_OBJECTS365 = {
+    "objects365_train": ("objects365/train", "objects365/annotations/objects365_train.json"),
+    "objects365_val": ("objects365/val", "objects365/annotations/objects365_val.json"),
+    "objects365_minival": ("objects365/val", "objects365/annotations/objects365_minival.json"),
+    "objects365_train_fixname": (
+        "objects365/train",
+        "objects365/annotations/objects365_train_fixname.json",
+    ),
+    "objects365_val_fixname": (
+        "objects365/val",
+        "objects365/annotations/objects365_val_fixname.json",
+    ),
+    "objects365_minival_fixname": (
+        "objects365/val",
+        "objects365/annotations/objects365_minival_fixname.json",
+    ),
+    "objects365_train_fixname_fixmiss": (
+        "objects365/train",
+        "objects365/annotations/objects365_train_fixname_fixmiss.json",
+    ),
+    "objects365_val_fixname_fixmiss": (
+        "objects365/val",
+        "objects365/annotations/objects365_val_fixname_fixmiss.json",
+    ),
+}
+
+
+def register_all_objects365(root):
+    for key, (image_root, json_file) in _PREDEFINED_SPLITS_OBJECTS365.items():
+        register_coco_instances(
+            key,
+            _get_builtin_metadata(key),
+            os.path.join(root, json_file) if "://" not in json_file else json_file,
+            os.path.join(root, image_root),
+        )
+
+
+if __name__.endswith(".objects365"):
+    # Assume pre-defined datasets live in `./datasets`.
+    _root = os.getenv("DETECTRON2_DATASETS", "datasets")
+    register_all_objects365(_root)

approach/ovod/APE/ape/data/datasets/odinw_categories.py

@@ -0,0 +1,377 @@
+ODINW_CATEGORIES = {
+    "AerialMaritimeDrone": [
+        {"id": 1, "name": "boat", "supercategory": "movable-objects"},
+        {"id": 2, "name": "car", "supercategory": "movable-objects"},
+        {"id": 3, "name": "dock", "supercategory": "movable-objects"},
+        {"id": 4, "name": "jetski", "supercategory": "movable-objects"},
+        {"id": 5, "name": "lift", "supercategory": "movable-objects"},
+    ],
+    "AmericanSignLanguageLetters": [
+        {"id": 1, "name": "A", "supercategory": "Letters"},
+        {"id": 2, "name": "B", "supercategory": "Letters"},
+        {"id": 3, "name": "C", "supercategory": "Letters"},
+        {"id": 4, "name": "D", "supercategory": "Letters"},
+        {"id": 5, "name": "E", "supercategory": "Letters"},
+        {"id": 6, "name": "F", "supercategory": "Letters"},
+        {"id": 7, "name": "G", "supercategory": "Letters"},
+        {"id": 8, "name": "H", "supercategory": "Letters"},
+        {"id": 9, "name": "I", "supercategory": "Letters"},
+        {"id": 10, "name": "J", "supercategory": "Letters"},
+        {"id": 11, "name": "K", "supercategory": "Letters"},
+        {"id": 12, "name": "L", "supercategory": "Letters"},
+        {"id": 13, "name": "M", "supercategory": "Letters"},
+        {"id": 14, "name": "N", "supercategory": "Letters"},
+        {"id": 15, "name": "O", "supercategory": "Letters"},
+        {"id": 16, "name": "P", "supercategory": "Letters"},
+        {"id": 17, "name": "Q", "supercategory": "Letters"},
+        {"id": 18, "name": "R", "supercategory": "Letters"},
+        {"id": 19, "name": "S", "supercategory": "Letters"},
+        {"id": 20, "name": "T", "supercategory": "Letters"},
+        {"id": 21, "name": "U", "supercategory": "Letters"},
+        {"id": 22, "name": "V", "supercategory": "Letters"},
+        {"id": 23, "name": "W", "supercategory": "Letters"},
+        {"id": 24, "name": "X", "supercategory": "Letters"},
+        {"id": 25, "name": "Y", "supercategory": "Letters"},
+        {"id": 26, "name": "Z", "supercategory": "Letters"},
+    ],
+    "Aquarium": [
+        {"id": 1, "name": "fish", "supercategory": "creatures"},
+        {"id": 2, "name": "jellyfish", "supercategory": "creatures"},
+        {"id": 3, "name": "penguin", "supercategory": "creatures"},
+        {"id": 4, "name": "puffin", "supercategory": "creatures"},
+        {"id": 5, "name": "shark", "supercategory": "creatures"},
+        {"id": 6, "name": "starfish", "supercategory": "creatures"},
+        {"id": 7, "name": "stingray", "supercategory": "creatures"},
+    ],
+    "BCCD": [
+        {"id": 1, "name": "Platelets", "supercategory": "cells"},
+        {"id": 2, "name": "RBC", "supercategory": "cells"},
+        {"id": 3, "name": "WBC", "supercategory": "cells"},
+    ],
+    "boggleBoards": [
+        {"id": 1, "name": "Q", "supercategory": "letters"},
+        {"id": 2, "name": "a", "supercategory": "letters"},
+        {"id": 3, "name": "an", "supercategory": "letters"},
+        {"id": 4, "name": "b", "supercategory": "letters"},
+        {"id": 5, "name": "c", "supercategory": "letters"},
+        {"id": 6, "name": "d", "supercategory": "letters"},
+        {"id": 7, "name": "e", "supercategory": "letters"},
+        {"id": 8, "name": "er", "supercategory": "letters"},
+        {"id": 9, "name": "f", "supercategory": "letters"},
+        {"id": 10, "name": "g", "supercategory": "letters"},
+        {"id": 11, "name": "h", "supercategory": "letters"},
+        {"id": 12, "name": "he", "supercategory": "letters"},
+        {"id": 13, "name": "i", "supercategory": "letters"},
+        {"id": 14, "name": "in", "supercategory": "letters"},
+        {"id": 15, "name": "j", "supercategory": "letters"},
+        {"id": 16, "name": "k", "supercategory": "letters"},
+        {"id": 17, "name": "l", "supercategory": "letters"},
+        {"id": 18, "name": "m", "supercategory": "letters"},
+        {"id": 19, "name": "n", "supercategory": "letters"},
+        {"id": 20, "name": "o", "supercategory": "letters"},
+        {"id": 21, "name": "o ", "supercategory": "letters"},
+        {"id": 22, "name": "p", "supercategory": "letters"},
+        {"id": 23, "name": "q", "supercategory": "letters"},
+        {"id": 24, "name": "qu", "supercategory": "letters"},
+        {"id": 25, "name": "r", "supercategory": "letters"},
+        {"id": 26, "name": "s", "supercategory": "letters"},
+        {"id": 27, "name": "t", "supercategory": "letters"},
+        {"id": 28, "name": "t\\", "supercategory": "letters"},
+        {"id": 29, "name": "th", "supercategory": "letters"},
+        {"id": 30, "name": "u", "supercategory": "letters"},
+        {"id": 31, "name": "v", "supercategory": "letters"},
+        {"id": 32, "name": "w", "supercategory": "letters"},
+        {"id": 33, "name": "wild", "supercategory": "letters"},
+        {"id": 34, "name": "x", "supercategory": "letters"},
+        {"id": 35, "name": "y", "supercategory": "letters"},
+        {"id": 36, "name": "z", "supercategory": "letters"},
+    ],
+    "brackishUnderwater": [
+        {"id": 1, "name": "crab", "supercategory": "animals"},
+        {"id": 2, "name": "fish", "supercategory": "animals"},
+        {"id": 3, "name": "jellyfish", "supercategory": "animals"},
+        {"id": 4, "name": "shrimp", "supercategory": "animals"},
+        {"id": 5, "name": "small_fish", "supercategory": "animals"},
+        {"id": 6, "name": "starfish", "supercategory": "animals"},
+    ],
+    "ChessPieces": [
+        {"id": 1, "name": "bishop", "supercategory": "pieces"},
+        {"id": 2, "name": "black-bishop", "supercategory": "pieces"},
+        {"id": 3, "name": "black-king", "supercategory": "pieces"},
+        {"id": 4, "name": "black-knight", "supercategory": "pieces"},
+        {"id": 5, "name": "black-pawn", "supercategory": "pieces"},
+        {"id": 6, "name": "black-queen", "supercategory": "pieces"},
+        {"id": 7, "name": "black-rook", "supercategory": "pieces"},
+        {"id": 8, "name": "white-bishop", "supercategory": "pieces"},
+        {"id": 9, "name": "white-king", "supercategory": "pieces"},
+        {"id": 10, "name": "white-knight", "supercategory": "pieces"},
+        {"id": 11, "name": "white-pawn", "supercategory": "pieces"},
+        {"id": 12, "name": "white-queen", "supercategory": "pieces"},
+        {"id": 13, "name": "white-rook", "supercategory": "pieces"},
+    ],
+    "CottontailRabbits": [
+        {"id": 1, "name": "Cottontail-Rabbit", "supercategory": "Cottontail-Rabbit"}
+    ],
+    "dice": [
+        {"id": 1, "name": "1", "supercategory": "dice"},
+        {"id": 2, "name": "2", "supercategory": "dice"},
+        {"id": 3, "name": "3", "supercategory": "dice"},
+        {"id": 4, "name": "4", "supercategory": "dice"},
+        {"id": 5, "name": "5", "supercategory": "dice"},
+        {"id": 6, "name": "6", "supercategory": "dice"},
+    ],
+    "DroneControl": [
+        {"id": 1, "name": "follow", "supercategory": "actions"},
+        {"id": 2, "name": "follow_hand", "supercategory": "actions"},
+        {"id": 3, "name": "land", "supercategory": "actions"},
+        {"id": 4, "name": "land_hand", "supercategory": "actions"},
+        {"id": 5, "name": "null", "supercategory": "actions"},
+        {"id": 6, "name": "object", "supercategory": "actions"},
+        {"id": 7, "name": "takeoff", "supercategory": "actions"},
+        {"id": 8, "name": "takeoff-hand", "supercategory": "actions"},
+    ],
+    "EgoHands-generic": [
+        {"id": 1, "name": "hand", "supercategory": "hands"},
+    ],
+    "EgoHands-specific": [
+        {"id": 1, "name": "myleft", "supercategory": "hands"},
+        {"id": 2, "name": "myright", "supercategory": "hands"},
+        {"id": 3, "name": "yourleft", "supercategory": "hands"},
+        {"id": 4, "name": "yourright", "supercategory": "hands"},
+    ],
+    "HardHatWorkers": [
+        {"id": 1, "name": "head", "supercategory": "Workers"},
+        {"id": 2, "name": "helmet", "supercategory": "Workers"},
+        {"id": 3, "name": "person", "supercategory": "Workers"},
+    ],
+    "MaskWearing": [
+        {"id": 1, "name": "mask", "supercategory": "People"},
+        {"id": 2, "name": "no-mask", "supercategory": "People"},
+    ],
+    "MountainDewCommercial": [
+        {"id": 1, "name": "bottle", "supercategory": "bottles"},
+    ],
+    "NorthAmericaMushrooms": [
+        {"id": 1, "name": "CoW", "supercategory": "mushroom"},
+        {"id": 2, "name": "chanterelle", "supercategory": "mushroom"},
+    ],
+    "openPoetryVision": [
+        {"id": 1, "name": "American Typewriter", "supercategory": "text"},
+        {"id": 2, "name": "Andale Mono", "supercategory": "text"},
+        {"id": 3, "name": "Apple Chancery", "supercategory": "text"},
+        {"id": 4, "name": "Arial", "supercategory": "text"},
+        {"id": 5, "name": "Avenir", "supercategory": "text"},
+        {"id": 6, "name": "Baskerville", "supercategory": "text"},
+        {"id": 7, "name": "Big Caslon", "supercategory": "text"},
+        {"id": 8, "name": "Bradley Hand", "supercategory": "text"},
+        {"id": 9, "name": "Brush Script MT", "supercategory": "text"},
+        {"id": 10, "name": "Chalkboard", "supercategory": "text"},
+        {"id": 11, "name": "Comic Sans MS", "supercategory": "text"},
+        {"id": 12, "name": "Copperplate", "supercategory": "text"},
+        {"id": 13, "name": "Courier", "supercategory": "text"},
+        {"id": 14, "name": "Didot", "supercategory": "text"},
+        {"id": 15, "name": "Futura", "supercategory": "text"},
+        {"id": 16, "name": "Geneva", "supercategory": "text"},
+        {"id": 17, "name": "Georgia", "supercategory": "text"},
+        {"id": 18, "name": "Gill Sans", "supercategory": "text"},
+        {"id": 19, "name": "Helvetica", "supercategory": "text"},
+        {"id": 20, "name": "Herculanum", "supercategory": "text"},
+        {"id": 21, "name": "Impact", "supercategory": "text"},
+        {"id": 22, "name": "Kefa", "supercategory": "text"},
+        {"id": 23, "name": "Lucida Grande", "supercategory": "text"},
+        {"id": 24, "name": "Luminari", "supercategory": "text"},
+        {"id": 25, "name": "Marker Felt", "supercategory": "text"},
+        {"id": 26, "name": "Menlo", "supercategory": "text"},
+        {"id": 27, "name": "Monaco", "supercategory": "text"},
+        {"id": 28, "name": "Noteworthy", "supercategory": "text"},
+        {"id": 29, "name": "Optima", "supercategory": "text"},
+        {"id": 30, "name": "PT Sans", "supercategory": "text"},
+        {"id": 31, "name": "PT Serif", "supercategory": "text"},
+        {"id": 32, "name": "Palatino", "supercategory": "text"},
+        {"id": 33, "name": "Papyrus", "supercategory": "text"},
+        {"id": 34, "name": "Phosphate", "supercategory": "text"},
+        {"id": 35, "name": "Rockwell", "supercategory": "text"},
+        {"id": 36, "name": "SF Pro", "supercategory": "text"},
+        {"id": 37, "name": "SignPainter", "supercategory": "text"},
+        {"id": 38, "name": "Skia", "supercategory": "text"},
+        {"id": 39, "name": "Snell Roundhand", "supercategory": "text"},
+        {"id": 40, "name": "Tahoma", "supercategory": "text"},
+        {"id": 41, "name": "Times New Roman", "supercategory": "text"},
+        {"id": 42, "name": "Trebuchet MS", "supercategory": "text"},
+        {"id": 43, "name": "Verdana", "supercategory": "text"},
+    ],
+    "OxfordPets-by-breed": [
+        {"id": 1, "name": "cat-Abyssinian", "supercategory": "pets"},
+        {"id": 2, "name": "cat-Bengal", "supercategory": "pets"},
+        {"id": 3, "name": "cat-Birman", "supercategory": "pets"},
+        {"id": 4, "name": "cat-Bombay", "supercategory": "pets"},
+        {"id": 5, "name": "cat-British_Shorthair", "supercategory": "pets"},
+        {"id": 6, "name": "cat-Egyptian_Mau", "supercategory": "pets"},
+        {"id": 7, "name": "cat-Maine_Coon", "supercategory": "pets"},
+        {"id": 8, "name": "cat-Persian", "supercategory": "pets"},
+        {"id": 9, "name": "cat-Ragdoll", "supercategory": "pets"},
+        {"id": 10, "name": "cat-Russian_Blue", "supercategory": "pets"},
+        {"id": 11, "name": "cat-Siamese", "supercategory": "pets"},
+        {"id": 12, "name": "cat-Sphynx", "supercategory": "pets"},
+        {"id": 13, "name": "dog-american_bulldog", "supercategory": "pets"},
+        {"id": 14, "name": "dog-american_pit_bull_terrier", "supercategory": "pets"},
+        {"id": 15, "name": "dog-basset_hound", "supercategory": "pets"},
+        {"id": 16, "name": "dog-beagle", "supercategory": "pets"},
+        {"id": 17, "name": "dog-boxer", "supercategory": "pets"},
+        {"id": 18, "name": "dog-chihuahua", "supercategory": "pets"},
+        {"id": 19, "name": "dog-english_cocker_spaniel", "supercategory": "pets"},
+        {"id": 20, "name": "dog-english_setter", "supercategory": "pets"},
+        {"id": 21, "name": "dog-german_shorthaired", "supercategory": "pets"},
+        {"id": 22, "name": "dog-great_pyrenees", "supercategory": "pets"},
+        {"id": 23, "name": "dog-havanese", "supercategory": "pets"},
+        {"id": 24, "name": "dog-japanese_chin", "supercategory": "pets"},
+        {"id": 25, "name": "dog-keeshond", "supercategory": "pets"},
+        {"id": 26, "name": "dog-leonberger", "supercategory": "pets"},
+        {"id": 27, "name": "dog-miniature_pinscher", "supercategory": "pets"},
+        {"id": 28, "name": "dog-newfoundland", "supercategory": "pets"},
+        {"id": 29, "name": "dog-pomeranian", "supercategory": "pets"},
+        {"id": 30, "name": "dog-pug", "supercategory": "pets"},
+        {"id": 31, "name": "dog-saint_bernard", "supercategory": "pets"},
+        {"id": 32, "name": "dog-samoyed", "supercategory": "pets"},
+        {"id": 33, "name": "dog-scottish_terrier", "supercategory": "pets"},
+        {"id": 34, "name": "dog-shiba_inu", "supercategory": "pets"},
+        {"id": 35, "name": "dog-staffordshire_bull_terrier", "supercategory": "pets"},
+        {"id": 36, "name": "dog-wheaten_terrier", "supercategory": "pets"},
+        {"id": 37, "name": "dog-yorkshire_terrier", "supercategory": "pets"},
+    ],
+    "OxfordPets-by-species": [
+        {"id": 1, "name": "cat", "supercategory": "pets"},
+        {"id": 2, "name": "dog", "supercategory": "pets"},
+    ],
+    "Packages": [{"id": 1, "name": "package", "supercategory": "packages"}],
+    "PascalVOC": [
+        {"id": 1, "name": "aeroplane", "supercategory": "VOC"},
+        {"id": 2, "name": "bicycle", "supercategory": "VOC"},
+        {"id": 3, "name": "bird", "supercategory": "VOC"},
+        {"id": 4, "name": "boat", "supercategory": "VOC"},
+        {"id": 5, "name": "bottle", "supercategory": "VOC"},
+        {"id": 6, "name": "bus", "supercategory": "VOC"},
+        {"id": 7, "name": "car", "supercategory": "VOC"},
+        {"id": 8, "name": "cat", "supercategory": "VOC"},
+        {"id": 9, "name": "chair", "supercategory": "VOC"},
+        {"id": 10, "name": "cow", "supercategory": "VOC"},
+        {"id": 11, "name": "diningtable", "supercategory": "VOC"},
+        {"id": 12, "name": "dog", "supercategory": "VOC"},
+        {"id": 13, "name": "horse", "supercategory": "VOC"},
+        {"id": 14, "name": "motorbike", "supercategory": "VOC"},
+        {"id": 15, "name": "person", "supercategory": "VOC"},
+        {"id": 16, "name": "pottedplant", "supercategory": "VOC"},
+        {"id": 17, "name": "sheep", "supercategory": "VOC"},
+        {"id": 18, "name": "sofa", "supercategory": "VOC"},
+        {"id": 19, "name": "train", "supercategory": "VOC"},
+        {"id": 20, "name": "tvmonitor", "supercategory": "VOC"},
+    ],
+    "pistols": [
+        {"id": 1, "name": "pistol", "supercategory": "Guns"},
+    ],
+    "PKLot": [
+        {"id": 1, "name": "space-empty", "supercategory": "spaces"},
+        {"id": 2, "name": "space-occupied", "supercategory": "spaces"},
+    ],
+    "plantdoc": [
+        {"id": 1, "name": "Apple Scab Leaf", "supercategory": "leaves"},
+        {"id": 2, "name": "Apple leaf", "supercategory": "leaves"},
+        {"id": 3, "name": "Apple rust leaf", "supercategory": "leaves"},
+        {"id": 4, "name": "Bell_pepper leaf", "supercategory": "leaves"},
+        {"id": 5, "name": "Bell_pepper leaf spot", "supercategory": "leaves"},
+        {"id": 6, "name": "Blueberry leaf", "supercategory": "leaves"},
+        {"id": 7, "name": "Cherry leaf", "supercategory": "leaves"},
+        {"id": 8, "name": "Corn Gray leaf spot", "supercategory": "leaves"},
+        {"id": 9, "name": "Corn leaf blight", "supercategory": "leaves"},
+        {"id": 10, "name": "Corn rust leaf", "supercategory": "leaves"},
+        {"id": 11, "name": "Peach leaf", "supercategory": "leaves"},
+        {"id": 12, "name": "Potato leaf", "supercategory": "leaves"},
+        {"id": 13, "name": "Potato leaf early blight", "supercategory": "leaves"},
+        {"id": 14, "name": "Potato leaf late blight", "supercategory": "leaves"},
+        {"id": 15, "name": "Raspberry leaf", "supercategory": "leaves"},
+        {"id": 16, "name": "Soyabean leaf", "supercategory": "leaves"},
+        {"id": 17, "name": "Soybean leaf", "supercategory": "leaves"},
+        {"id": 18, "name": "Squash Powdery mildew leaf", "supercategory": "leaves"},
+        {"id": 19, "name": "Strawberry leaf", "supercategory": "leaves"},
+        {"id": 20, "name": "Tomato Early blight leaf", "supercategory": "leaves"},
+        {"id": 21, "name": "Tomato Septoria leaf spot", "supercategory": "leaves"},
+        {"id": 22, "name": "Tomato leaf", "supercategory": "leaves"},
+        {"id": 23, "name": "Tomato leaf bacterial spot", "supercategory": "leaves"},
+        {"id": 24, "name": "Tomato leaf late blight", "supercategory": "leaves"},
+        {"id": 25, "name": "Tomato leaf mosaic virus", "supercategory": "leaves"},
+        {"id": 26, "name": "Tomato leaf yellow virus", "supercategory": "leaves"},
+        {"id": 27, "name": "Tomato mold leaf", "supercategory": "leaves"},
+        {"id": 28, "name": "Tomato two spotted spider mites leaf", "supercategory": "leaves"},
+        {"id": 29, "name": "grape leaf", "supercategory": "leaves"},
+        {"id": 30, "name": "grape leaf black rot", "supercategory": "leaves"},
+    ],
+    "pothole": [
+        {"id": 1, "name": "pothole", "supercategory": "potholes"},
+    ],
+    "Raccoon": [
+        {"id": 1, "name": "raccoon", "supercategory": "raccoons"},
+    ],
+    "selfdrivingCar": [
+        {"id": 1, "name": "biker", "supercategory": "obstacles"},
+        {"id": 2, "name": "car", "supercategory": "obstacles"},
+        {"id": 3, "name": "pedestrian", "supercategory": "obstacles"},
+        {"id": 4, "name": "trafficLight", "supercategory": "obstacles"},
+        {"id": 5, "name": "trafficLight-Green", "supercategory": "obstacles"},
+        {"id": 6, "name": "trafficLight-GreenLeft", "supercategory": "obstacles"},
+        {"id": 7, "name": "trafficLight-Red", "supercategory": "obstacles"},
+        {"id": 8, "name": "trafficLight-RedLeft", "supercategory": "obstacles"},
+        {"id": 9, "name": "trafficLight-Yellow", "supercategory": "obstacles"},
+        {"id": 10, "name": "trafficLight-YellowLeft", "supercategory": "obstacles"},
+        {"id": 11, "name": "truck", "supercategory": "obstacles"},
+    ],
+    "ShellfishOpenImages": [
+        {"id": 1, "name": "Crab", "supercategory": "shellfish"},
+        {"id": 2, "name": "Lobster", "supercategory": "shellfish"},
+        {"id": 3, "name": "Shrimp", "supercategory": "shellfish"},
+    ],
+    "ThermalCheetah": [
+        {"id": 1, "name": "cheetah", "supercategory": "cheetah"},
+        {"id": 2, "name": "human", "supercategory": "cheetah"},
+    ],
+    "thermalDogsAndPeople": [
+        {"id": 1, "name": "dog", "supercategory": "dogs-person"},
+        {"id": 2, "name": "person", "supercategory": "dogs-person"},
+    ],
+    "UnoCards": [
+        {"id": 1, "name": "0", "supercategory": "Card-Types"},
+        {"id": 2, "name": "1", "supercategory": "Card-Types"},
+        {"id": 3, "name": "2", "supercategory": "Card-Types"},
+        {"id": 4, "name": "3", "supercategory": "Card-Types"},
+        {"id": 5, "name": "4", "supercategory": "Card-Types"},
+        {"id": 6, "name": "5", "supercategory": "Card-Types"},
+        {"id": 7, "name": "6", "supercategory": "Card-Types"},
+        {"id": 8, "name": "7", "supercategory": "Card-Types"},
+        {"id": 9, "name": "8", "supercategory": "Card-Types"},
+        {"id": 10, "name": "9", "supercategory": "Card-Types"},
+        {"id": 11, "name": "10", "supercategory": "Card-Types"},
+        {"id": 12, "name": "11", "supercategory": "Card-Types"},
+        {"id": 13, "name": "12", "supercategory": "Card-Types"},
+        {"id": 14, "name": "13", "supercategory": "Card-Types"},
+        {"id": 15, "name": "14", "supercategory": "Card-Types"},
+    ],
+    "VehiclesOpenImages": [
+        {"id": 1, "name": "Ambulance", "supercategory": "vehicles"},
+        {"id": 2, "name": "Bus", "supercategory": "vehicles"},
+        {"id": 3, "name": "Car", "supercategory": "vehicles"},
+        {"id": 4, "name": "Motorcycle", "supercategory": "vehicles"},
+        {"id": 5, "name": "Truck", "supercategory": "vehicles"},
+    ],
+    "websiteScreenshots": [
+        {"id": 1, "name": "button", "supercategory": "elements"},
+        {"id": 2, "name": "field", "supercategory": "elements"},
+        {"id": 3, "name": "heading", "supercategory": "elements"},
+        {"id": 4, "name": "iframe", "supercategory": "elements"},
+        {"id": 5, "name": "image", "supercategory": "elements"},
+        {"id": 6, "name": "label", "supercategory": "elements"},
+        {"id": 7, "name": "link", "supercategory": "elements"},
+        {"id": 8, "name": "text", "supercategory": "elements"},
+    ],
+    "WildfireSmoke": [
+        {"id": 1, "name": "smoke", "supercategory": "Smoke"},
+    ],
+}

approach/ovod/APE/ape/data/datasets/odinw_instance.py

@@ -0,0 +1,824 @@
+import contextlib
+import io
+import logging
+import os
+
+import pycocotools.mask as mask_util
+
+from detectron2.data import DatasetCatalog, MetadataCatalog
+from detectron2.structures import BoxMode
+from detectron2.utils.file_io import PathManager
+from fvcore.common.timer import Timer
+
+from .odinw_categories import ODINW_CATEGORIES
+from .odinw_prompts import ODINW_PROMPTS
+
+logger = logging.getLogger(__name__)
+
+
+def load_coco_json(json_file, image_root, dataset_name=None, extra_annotation_keys=None):
+    """
+    Load a json file with COCO's instances annotation format.
+    Currently supports instance detection, instance segmentation,
+    and person keypoints annotations.
+
+    Args:
+        json_file (str): full path to the json file in COCO instances annotation format.
+        image_root (str or path-like): the directory where the images in this json file exists.
+        dataset_name (str or None): the name of the dataset (e.g., coco_2017_train).
+            When provided, this function will also do the following:
+
+            * Put "thing_classes" into the metadata associated with this dataset.
+            * Map the category ids into a contiguous range (needed by standard dataset format),
+              and add "thing_dataset_id_to_contiguous_id" to the metadata associated
+              with this dataset.
+
+            This option should usually be provided, unless users need to load
+            the original json content and apply more processing manually.
+        extra_annotation_keys (list[str]): list of per-annotation keys that should also be
+            loaded into the dataset dict (besides "iscrowd", "bbox", "keypoints",
+            "category_id", "segmentation"). The values for these keys will be returned as-is.
+            For example, the densepose annotations are loaded in this way.
+
+    Returns:
+        list[dict]: a list of dicts in Detectron2 standard dataset dicts format (See
+        `Using Custom Datasets </tutorials/datasets.html>`_ ) when `dataset_name` is not None.
+        If `dataset_name` is None, the returned `category_ids` may be
+        incontiguous and may not conform to the Detectron2 standard format.
+
+    Notes:
+        1. This function does not read the image files.
+           The results do not have the "image" field.
+    """
+    from pycocotools.coco import COCO
+
+    timer = Timer()
+    json_file = PathManager.get_local_path(json_file)
+    with contextlib.redirect_stdout(io.StringIO()):
+        coco_api = COCO(json_file)
+    if timer.seconds() > 1:
+        logger.info("Loading {} takes {:.2f} seconds.".format(json_file, timer.seconds()))
+
+    id_map = None
+    if dataset_name is not None:
+        meta = MetadataCatalog.get(dataset_name)
+        cat_ids = sorted(coco_api.getCatIds())
+        # cats = coco_api.loadCats(cat_ids)
+        # The categories in a custom json file may not be sorted.
+        # thing_classes = [c["name"] for c in sorted(cats, key=lambda x: x["id"])]
+        # meta.thing_classes = thing_classes
+
+        # In COCO, certain category ids are artificially removed,
+        # and by convention they are always ignored.
+        # We deal with COCO's id issue and translate
+        # the category ids to contiguous ids in [0, 80).
+
+        # It works by looking at the "categories" field in the json, therefore
+        # if users' own json also have incontiguous ids, we'll
+        # apply this mapping as well but print a warning.
+        if not (min(cat_ids) == 1 and max(cat_ids) == len(cat_ids)):
+            if "coco" not in dataset_name:
+                logger.warning(
+                    """
+Category ids in annotations are not in [1, #categories]! We'll apply a mapping for you.
+"""
+                )
+        id_map = {v: i for i, v in enumerate(cat_ids)}
+        meta.thing_dataset_id_to_contiguous_id = id_map
+
+    # sort indices for reproducible results
+    img_ids = sorted(coco_api.imgs.keys())
+    # imgs is a list of dicts, each looks something like:
+    # {'license': 4,
+    #  'url': 'http://farm6.staticflickr.com/5454/9413846304_881d5e5c3b_z.jpg',
+    #  'file_name': 'COCO_val2014_000000001268.jpg',
+    #  'height': 427,
+    #  'width': 640,
+    #  'date_captured': '2013-11-17 05:57:24',
+    #  'id': 1268}
+    imgs = coco_api.loadImgs(img_ids)
+    # anns is a list[list[dict]], where each dict is an annotation
+    # record for an object. The inner list enumerates the objects in an image
+    # and the outer list enumerates over images. Example of anns[0]:
+    # [{'segmentation': [[192.81,
+    #     247.09,
+    #     ...
+    #     219.03,
+    #     249.06]],
+    #   'area': 1035.749,
+    #   'iscrowd': 0,
+    #   'image_id': 1268,
+    #   'bbox': [192.81, 224.8, 74.73, 33.43],
+    #   'category_id': 16,
+    #   'id': 42986},
+    #  ...]
+    anns = [coco_api.imgToAnns[img_id] for img_id in img_ids]
+    total_num_valid_anns = sum([len(x) for x in anns])
+    total_num_anns = len(coco_api.anns)
+    if total_num_valid_anns < total_num_anns:
+        logger.warning(
+            f"{json_file} contains {total_num_anns} annotations, but only "
+            f"{total_num_valid_anns} of them match to images in the file."
+        )
+
+    if "minival" not in json_file:
+        # The popular valminusminival & minival annotations for COCO2014 contain this bug.
+        # However the ratio of buggy annotations there is tiny and does not affect accuracy.
+        # Therefore we explicitly white-list them.
+        ann_ids = [ann["id"] for anns_per_image in anns for ann in anns_per_image]
+        assert len(set(ann_ids)) == len(ann_ids), "Annotation ids in '{}' are not unique!".format(
+            json_file
+        )
+
+    imgs_anns = list(zip(imgs, anns))
+    logger.info("Loaded {} images in COCO format from {}".format(len(imgs_anns), json_file))
+
+    dataset_dicts = []
+
+    ann_keys = ["iscrowd", "bbox", "keypoints", "category_id"] + (extra_annotation_keys or [])
+
+    num_instances_without_valid_segmentation = 0
+
+    for (img_dict, anno_dict_list) in imgs_anns:
+        record = {}
+        record["file_name"] = os.path.join(image_root, img_dict["file_name"])
+        # record["height"] = img_dict["height"]
+        # record["width"] = img_dict["width"]
+        image_id = record["image_id"] = img_dict["id"]
+
+        objs = []
+        for anno in anno_dict_list:
+            # Check that the image_id in this annotation is the same as
+            # the image_id we're looking at.
+            # This fails only when the data parsing logic or the annotation file is buggy.
+
+            # The original COCO valminusminival2014 & minival2014 annotation files
+            # actually contains bugs that, together with certain ways of using COCO API,
+            # can trigger this assertion.
+            assert anno["image_id"] == image_id
+
+            assert anno.get("ignore", 0) == 0, '"ignore" in COCO json file is not supported.'
+
+            obj = {key: anno[key] for key in ann_keys if key in anno}
+            if "bbox" in obj and len(obj["bbox"]) == 0:
+                raise ValueError(
+                    f"One annotation of image {image_id} contains empty 'bbox' value! "
+                    "This json does not have valid COCO format."
+                )
+
+            segm = anno.get("segmentation", None)
+            if segm:  # either list[list[float]] or dict(RLE)
+                if isinstance(segm, dict):
+                    if isinstance(segm["counts"], list):
+                        # convert to compressed RLE
+                        segm = mask_util.frPyObjects(segm, *segm["size"])
+                else:
+                    # filter out invalid polygons (< 3 points)
+                    segm = [poly for poly in segm if len(poly) % 2 == 0 and len(poly) >= 6]
+                    if len(segm) == 0:
+                        num_instances_without_valid_segmentation += 1
+                        continue  # ignore this instance
+                obj["segmentation"] = segm
+
+            keypts = anno.get("keypoints", None)
+            if keypts:  # list[int]
+                for idx, v in enumerate(keypts):
+                    if idx % 3 != 2:
+                        # COCO's segmentation coordinates are floating points in [0, H or W],
+                        # but keypoint coordinates are integers in [0, H-1 or W-1]
+                        # Therefore we assume the coordinates are "pixel indices" and
+                        # add 0.5 to convert to floating point coordinates.
+                        keypts[idx] = v + 0.5
+                obj["keypoints"] = keypts
+
+            obj["bbox_mode"] = BoxMode.XYWH_ABS
+            if id_map:
+                annotation_category_id = obj["category_id"]
+                try:
+                    obj["category_id"] = id_map[annotation_category_id]
+                except KeyError as e:
+                    raise KeyError(
+                        f"Encountered category_id={annotation_category_id} "
+                        "but this id does not exist in 'categories' of the json file."
+                    ) from e
+            objs.append(obj)
+        record["annotations"] = objs
+        dataset_dicts.append(record)
+
+    if num_instances_without_valid_segmentation > 0:
+        logger.warning(
+            "Filtered out {} instances without valid segmentation. ".format(
+                num_instances_without_valid_segmentation
+            )
+            + "There might be issues in your dataset generation process.  Please "
+            "check https://detectron2.readthedocs.io/en/latest/tutorials/datasets.html carefully"
+        )
+    return dataset_dicts
+
+
+def register_coco_instances(name, metadata, json_file, image_root):
+    """
+    Register a dataset in COCO's json annotation format for
+    instance detection, instance segmentation and keypoint detection.
+    (i.e., Type 1 and 2 in http://cocodataset.org/#format-data.
+    `instances*.json` and `person_keypoints*.json` in the dataset).
+
+    This is an example of how to register a new dataset.
+    You can do something similar to this function, to register new datasets.
+
+    Args:
+        name (str): the name that identifies a dataset, e.g. "coco_2014_train".
+        metadata (dict): extra metadata associated with this dataset.  You can
+            leave it as an empty dict.
+        json_file (str): path to the json instance annotation file.
+        image_root (str or path-like): directory which contains all the images.
+    """
+    assert isinstance(name, str), name
+    assert isinstance(json_file, (str, os.PathLike)), json_file
+    assert isinstance(image_root, (str, os.PathLike)), image_root
+    # 1. register a function which returns dicts
+    DatasetCatalog.register(name, lambda: load_coco_json(json_file, image_root, name))
+
+    # 2. Optionally, add metadata about this dataset,
+    # since they might be useful in evaluation, visualization or logging
+    MetadataCatalog.get(name).set(
+        json_file=json_file, image_root=image_root, evaluator_type="coco", **metadata
+    )
+
+
+_PREDEFINED_SPLITS_ODINW = {
+    "odinw_AerialMaritimeDrone_large": {
+        "odinw_AerialMaritimeDrone_large_train": (
+            "odinw/AerialMaritimeDrone/large/train/",
+            "odinw/AerialMaritimeDrone/large/train/annotations_without_background_converted.json",
+        ),
+        "odinw_AerialMaritimeDrone_large_val": (
+            "odinw/AerialMaritimeDrone/large/valid/",
+            "odinw/AerialMaritimeDrone/large/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_AerialMaritimeDrone_large_test": (
+            "odinw/AerialMaritimeDrone/large/test/",
+            "odinw/AerialMaritimeDrone/large/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_AerialMaritimeDrone_tiled": {
+        "odinw_AerialMaritimeDrone_tiled_train": (
+            "odinw/AerialMaritimeDrone/tiled/train/",
+            "odinw/AerialMaritimeDrone/tiled/train/annotations_without_background_converted.json",
+        ),
+        "odinw_AerialMaritimeDrone_tiled_val": (
+            "odinw/AerialMaritimeDrone/tiled/valid/",
+            "odinw/AerialMaritimeDrone/tiled/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_AerialMaritimeDrone_tiled_test": (
+            "odinw/AerialMaritimeDrone/tiled/test/",
+            "odinw/AerialMaritimeDrone/tiled/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_AmericanSignLanguageLetters_American_Sign_Language_Letters.v1-v1.coco": {
+        "odinw_AmericanSignLanguageLetters_American_Sign_Language_Letters.v1-v1.coco_train": (
+            "odinw/AmericanSignLanguageLetters/American Sign Language Letters.v1-v1.coco/train/",
+            "odinw/AmericanSignLanguageLetters/American Sign Language Letters.v1-v1.coco/train/annotations_without_background_converted.json",
+        ),
+        "odinw_AmericanSignLanguageLetters_American_Sign_Language_Letters.v1-v1.coco_val": (
+            "odinw/AmericanSignLanguageLetters/American Sign Language Letters.v1-v1.coco/valid/",
+            "odinw/AmericanSignLanguageLetters/American Sign Language Letters.v1-v1.coco/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_AmericanSignLanguageLetters_American_Sign_Language_Letters.v1-v1.coco_test": (
+            "odinw/AmericanSignLanguageLetters/American Sign Language Letters.v1-v1.coco/test/",
+            "odinw/AmericanSignLanguageLetters/American Sign Language Letters.v1-v1.coco/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_Aquarium_Aquarium_Combined.v2-raw-1024.coco": {
+        "odinw_Aquarium_Aquarium_Combined.v2-raw-1024.coco_train": (
+            "odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/",
+            "odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/annotations_without_background_converted.json",
+        ),
+        "odinw_Aquarium_Aquarium_Combined.v2-raw-1024.coco_val": (
+            "odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/",
+            "odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_Aquarium_Aquarium_Combined.v2-raw-1024.coco_test": (
+            "odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/test/",
+            "odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_BCCD_BCCD.v3-raw.coco": {
+        "odinw_BCCD_BCCD.v3-raw.coco_train": (
+            "odinw/BCCD/BCCD.v3-raw.coco/train/",
+            "odinw/BCCD/BCCD.v3-raw.coco/train/annotations_without_background_converted.json",
+        ),
+        "odinw_BCCD_BCCD.v3-raw.coco_val": (
+            "odinw/BCCD/BCCD.v3-raw.coco/valid/",
+            "odinw/BCCD/BCCD.v3-raw.coco/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_BCCD_BCCD.v3-raw.coco_test": (
+            "odinw/BCCD/BCCD.v3-raw.coco/test/",
+            "odinw/BCCD/BCCD.v3-raw.coco/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_boggleBoards_416x416AutoOrient_export_": {
+        "odinw_boggleBoards_416x416AutoOrient_export_train": (
+            "odinw/boggleBoards/416x416AutoOrient/export/",
+            "odinw/boggleBoards/416x416AutoOrient/export/train_annotations_without_background_converted.json",
+        ),
+        "odinw_boggleBoards_416x416AutoOrient_export_val": (
+            "odinw/boggleBoards/416x416AutoOrient/export/",
+            "odinw/boggleBoards/416x416AutoOrient/export/val_annotations_without_background_converted.json",
+        ),
+        "odinw_boggleBoards_416x416AutoOrient_export_test": (
+            "odinw/boggleBoards/416x416AutoOrient/export/",
+            "odinw/boggleBoards/416x416AutoOrient/export/test_annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_brackishUnderwater_960x540": {
+        "odinw_brackishUnderwater_960x540_train": (
+            "odinw/brackishUnderwater/960x540/train/",
+            "odinw/brackishUnderwater/960x540/train/annotations_without_background_converted.json",
+        ),
+        "odinw_brackishUnderwater_960x540_val": (
+            "odinw/brackishUnderwater/960x540/valid/",
+            "odinw/brackishUnderwater/960x540/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_brackishUnderwater_960x540_minival": (
+            "odinw/brackishUnderwater/960x540/mini_val/",
+            "odinw/brackishUnderwater/960x540/mini_val/annotations_without_background_converted.json",
+        ),
+        "odinw_brackishUnderwater_960x540_test": (
+            "odinw/brackishUnderwater/960x540/test/",
+            "odinw/brackishUnderwater/960x540/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_ChessPieces_Chess_Pieces.v23-raw.coco": {
+        "odinw_ChessPieces_Chess_Pieces.v23-raw.coco_train": (
+            "odinw/ChessPieces/Chess Pieces.v23-raw.coco/train/",
+            "odinw/ChessPieces/Chess Pieces.v23-raw.coco/train/annotations_without_background_converted.json",
+        ),
+        "odinw_ChessPieces_Chess_Pieces.v23-raw.coco_val": (
+            "odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid/",
+            "odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_ChessPieces_Chess_Pieces.v23-raw.coco_test": (
+            "odinw/ChessPieces/Chess Pieces.v23-raw.coco/test/",
+            "odinw/ChessPieces/Chess Pieces.v23-raw.coco/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_CottontailRabbits": {
+        "odinw_CottontailRabbits_train": (
+            "odinw/CottontailRabbits/train/",
+            "odinw/CottontailRabbits/train/annotations_without_background_converted.json",
+        ),
+        "odinw_CottontailRabbits_val": (
+            "odinw/CottontailRabbits/valid/",
+            "odinw/CottontailRabbits/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_CottontailRabbits_test": (
+            "odinw/CottontailRabbits/test/",
+            "odinw/CottontailRabbits/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_dice_mediumColor_export": {
+        "odinw_dice_mediumColor_export_train": (
+            "odinw/dice/mediumColor/export/",
+            "odinw/dice/mediumColor/export/train_annotations_without_background_converted.json",
+        ),
+        "odinw_dice_mediumColor_export_val": (
+            "odinw/dice/mediumColor/export/",
+            "odinw/dice/mediumColor/export/val_annotations_without_background_converted.json",
+        ),
+        "odinw_dice_mediumColor_export_test": (
+            "odinw/dice/mediumColor/export/",
+            "odinw/dice/mediumColor/export/test_annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_DroneControl_Drone_Control.v3-raw.coco": {
+        "odinw_DroneControl_Drone_Control.v3-raw.coco_train": (
+            "odinw/DroneControl/Drone Control.v3-raw.coco/train/",
+            "odinw/DroneControl/Drone Control.v3-raw.coco/train/annotations_without_background_converted.json",
+        ),
+        "odinw_DroneControl_Drone_Control.v3-raw.coco_val": (
+            "odinw/DroneControl/Drone Control.v3-raw.coco/valid/",
+            "odinw/DroneControl/Drone Control.v3-raw.coco/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_DroneControl_Drone_Control.v3-raw.coco_minival": (
+            "odinw/DroneControl/Drone Control.v3-raw.coco/mini_val/",
+            "odinw/DroneControl/Drone Control.v3-raw.coco/mini_val/annotations_without_background_converted.json",
+        ),
+        "odinw_DroneControl_Drone_Control.v3-raw.coco_test": (
+            "odinw/DroneControl/Drone Control.v3-raw.coco/test/",
+            "odinw/DroneControl/Drone Control.v3-raw.coco/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_EgoHands-generic": {
+        "odinw_EgoHands_generic_train": (
+            "odinw/EgoHands/generic/train/",
+            "odinw/EgoHands/generic/train/annotations_without_background_converted.json",
+        ),
+        "odinw_EgoHands_generic_val": (
+            "odinw/EgoHands/generic/valid/",
+            "odinw/EgoHands/generic/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_EgoHands_generic_minival": (
+            "odinw/EgoHands/generic/mini_val/",
+            "odinw/EgoHands/generic/mini_val/annotations_without_background_converted.json",
+        ),
+        "odinw_EgoHands_generic_test": (
+            "odinw/EgoHands/generic/test/",
+            "odinw/EgoHands/generic/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_EgoHands-specific": {
+        "odinw_EgoHands_specific_train": (
+            "odinw/EgoHands/specific/train/",
+            "odinw/EgoHands/specific/train/annotations_without_background_converted.json",
+        ),
+        "odinw_EgoHands_specific_val": (
+            "odinw/EgoHands/specific/valid/",
+            "odinw/EgoHands/specific/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_EgoHands_specific_minival": (
+            "odinw/EgoHands/specific/mini_val/",
+            "odinw/EgoHands/specific/mini_val/annotations_without_background_converted.json",
+        ),
+        "odinw_EgoHands_specific_test": (
+            "odinw/EgoHands/specific/test/",
+            "odinw/EgoHands/specific/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_HardHatWorkers_raw": {
+        "odinw_HardHatWorkers_raw_train": (
+            "odinw/HardHatWorkers/raw/train/",
+            "odinw/HardHatWorkers/raw/train/annotations_without_background_converted.json",
+        ),
+        "odinw_HardHatWorkers_raw_val": (
+            "odinw/HardHatWorkers/raw/valid/",
+            "odinw/HardHatWorkers/raw/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_HardHatWorkers_raw_test": (
+            "odinw/HardHatWorkers/raw/test/",
+            "odinw/HardHatWorkers/raw/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_MaskWearing_raw": {
+        "odinw_MaskWearing_raw_train": (
+            "odinw/MaskWearing/raw/train/",
+            "odinw/MaskWearing/raw/train/annotations_without_background_converted.json",
+        ),
+        "odinw_MaskWearing_raw_val": (
+            "odinw/MaskWearing/raw/valid/",
+            "odinw/MaskWearing/raw/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_MaskWearing_raw_test": (
+            "odinw/MaskWearing/raw/test/",
+            "odinw/MaskWearing/raw/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_MountainDewCommercial": {
+        "odinw_MountainDewCommercial_train": (
+            "odinw/MountainDewCommercial/train/",
+            "odinw/MountainDewCommercial/train/annotations_without_background_converted.json",
+        ),
+        "odinw_MountainDewCommercial_val": (
+            "odinw/MountainDewCommercial/valid/",
+            "odinw/MountainDewCommercial/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_MountainDewCommercial_test": (
+            "odinw/MountainDewCommercial/test/",
+            "odinw/MountainDewCommercial/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_NorthAmericaMushrooms_North_American_Mushrooms.v1-416x416.coco": {
+        "odinw_NorthAmericaMushrooms_North_American_Mushrooms.v1-416x416.coco_train": (
+            "odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/",
+            "odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/annotations_without_background_converted.json",
+        ),
+        "odinw_NorthAmericaMushrooms_North_American_Mushrooms.v1-416x416.coco_val": (
+            "odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/",
+            "odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_NorthAmericaMushrooms_North_American_Mushrooms.v1-416x416.coco_test": (
+            "odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/test/",
+            "odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_openPoetryVision_512x512": {
+        "odinw_openPoetryVision_512x512_train": (
+            "odinw/openPoetryVision/512x512/train/",
+            "odinw/openPoetryVision/512x512/train/annotations_without_background_converted.json",
+        ),
+        "odinw_openPoetryVision_512x512_val": (
+            "odinw/openPoetryVision/512x512/valid/",
+            "odinw/openPoetryVision/512x512/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_openPoetryVision_512x512_minival": (
+            "odinw/openPoetryVision/512x512/mini_val/",
+            "odinw/openPoetryVision/512x512/mini_val/annotations_without_background_converted.json",
+        ),
+        "odinw_openPoetryVision_512x512_test": (
+            "odinw/openPoetryVision/512x512/test/",
+            "odinw/openPoetryVision/512x512/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_OxfordPets-by-breed": {
+        "odinw_OxfordPets_by-breed_train": (
+            "odinw/OxfordPets/by-breed/train/",
+            "odinw/OxfordPets/by-breed/train/annotations_without_background_converted.json",
+        ),
+        "odinw_OxfordPets_by-breed_val": (
+            "odinw/OxfordPets/by-breed/valid/",
+            "odinw/OxfordPets/by-breed/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_OxfordPets_by-breed_minival": (
+            "odinw/OxfordPets/by-breed/mini_val/",
+            "odinw/OxfordPets/by-breed/mini_val/annotations_without_background_converted.json",
+        ),
+        "odinw_OxfordPets_by-breed_test": (
+            "odinw/OxfordPets/by-breed/test/",
+            "odinw/OxfordPets/by-breed/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_OxfordPets-by-species": {
+        "odinw_OxfordPets_by-species_train": (
+            "odinw/OxfordPets/by-species/train/",
+            "odinw/OxfordPets/by-species/train/annotations_without_background_converted.json",
+        ),
+        "odinw_OxfordPets_by-species_val": (
+            "odinw/OxfordPets/by-species/valid/",
+            "odinw/OxfordPets/by-species/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_OxfordPets_by-species_minival": (
+            "odinw/OxfordPets/by-species/mini_val/",
+            "odinw/OxfordPets/by-species/mini_val/annotations_without_background_converted.json",
+        ),
+        "odinw_OxfordPets_by-species_test": (
+            "odinw/OxfordPets/by-species/test/",
+            "odinw/OxfordPets/by-species/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_Packages_Raw": {
+        "odinw_Packages_Raw_train": (
+            "odinw/Packages/Raw/train/",
+            "odinw/Packages/Raw/train/annotations_without_background_converted.json",
+        ),
+        "odinw_Packages_Raw_val": (
+            "odinw/Packages/Raw/valid/",
+            "odinw/Packages/Raw/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_Packages_Raw_test": (
+            "odinw/Packages/Raw/test/",
+            "odinw/Packages/Raw/test/annotations_without_background_converted.json",
+            # "odinw/Packages/Raw/test/_annotations.coco_converted.json",
+        ),
+    },
+    "odinw_PascalVOC": {
+        "odinw_PascalVOC_train": (
+            "odinw/PascalVOC/train/",
+            "odinw/PascalVOC/train/annotations_without_background_converted.json",
+        ),
+        "odinw_PascalVOC_val": (
+            "odinw/PascalVOC/valid/",
+            "odinw/PascalVOC/valid/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_pistols_export": {
+        "odinw_pistols_export_train": (
+            "odinw/pistols/export/",
+            "odinw/pistols/export/train_annotations_without_background_converted.json",
+        ),
+        "odinw_pistols_export_val": (
+            "odinw/pistols/export/",
+            "odinw/pistols/export/val_annotations_without_background_converted.json",
+        ),
+        "odinw_pistols_export_test": (
+            "odinw/pistols/export/",
+            "odinw/pistols/export/test_annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_PKLot_640": {
+        "odinw_PKLot_640_train": (
+            "odinw/PKLot/640/train/",
+            "odinw/PKLot/640/train/annotations_without_background_converted.json",
+        ),
+        "odinw_PKLot_640_val": (
+            "odinw/PKLot/640/valid/",
+            "odinw/PKLot/640/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_PKLot_640_minival": (
+            "odinw/PKLot/640/mini_val/",
+            "odinw/PKLot/640/mini_val/annotations_without_background_converted.json",
+        ),
+        "odinw_PKLot_640_test": (
+            "odinw/PKLot/640/test/",
+            "odinw/PKLot/640/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_plantdoc_100x100": {
+        "odinw_plantdoc_100x100_train": (
+            "odinw/plantdoc/100x100/train/",
+            "odinw/plantdoc/100x100/train/annotations_without_background_converted.json",
+        ),
+        "odinw_plantdoc_100x100_val": (
+            "odinw/plantdoc/100x100/valid/",
+            "odinw/plantdoc/100x100/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_plantdoc_100x100_test": (
+            "odinw/plantdoc/100x100/test/",
+            "odinw/plantdoc/100x100/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_plantdoc_416x416": {
+        "odinw_plantdoc_416x416_train": (
+            "odinw/plantdoc/416x416/train/",
+            "odinw/plantdoc/416x416/train/annotations_without_background_converted.json",
+        ),
+        "odinw_plantdoc_416x416_val": (
+            "odinw/plantdoc/416x416/valid/",
+            "odinw/plantdoc/416x416/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_plantdoc_416x416_test": (
+            "odinw/plantdoc/416x416/test/",
+            "odinw/plantdoc/416x416/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_pothole": {
+        "odinw_pothole_train": (
+            "odinw/pothole/train/",
+            "odinw/pothole/train/annotations_without_background_converted.json",
+        ),
+        "odinw_pothole_val": (
+            "odinw/pothole/valid/",
+            "odinw/pothole/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_pothole_test": (
+            "odinw/pothole/test/",
+            "odinw/pothole/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_Raccoon_Raccoon.v2-raw.coco": {
+        "odinw_Raccoon_Raccoon.v2-raw.coco_train": (
+            "odinw/Raccoon/Raccoon.v2-raw.coco/train/",
+            "odinw/Raccoon/Raccoon.v2-raw.coco/train/annotations_without_background_converted.json",
+        ),
+        "odinw_Raccoon_Raccoon.v2-raw.coco_val": (
+            "odinw/Raccoon/Raccoon.v2-raw.coco/valid/",
+            "odinw/Raccoon/Raccoon.v2-raw.coco/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_Raccoon_Raccoon.v2-raw.coco_test": (
+            "odinw/Raccoon/Raccoon.v2-raw.coco/test/",
+            "odinw/Raccoon/Raccoon.v2-raw.coco/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_selfdrivingCar_fixedLarge_export": {
+        "odinw_selfdrivingCar_fixedLarge_export_train": (
+            "odinw/selfdrivingCar/fixedLarge/export/",
+            "odinw/selfdrivingCar/fixedLarge/export/train_annotations_without_background_converted.json",
+        ),
+        "odinw_selfdrivingCar_fixedLarge_export_val": (
+            "odinw/selfdrivingCar/fixedLarge/export/",
+            "odinw/selfdrivingCar/fixedLarge/export/val_annotations_without_background_converted.json",
+        ),
+        "odinw_selfdrivingCar_fixedLarge_export_test": (
+            "odinw/selfdrivingCar/fixedLarge/export/",
+            "odinw/selfdrivingCar/fixedLarge/export/test_annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_ShellfishOpenImages_raw": {
+        "odinw_ShellfishOpenImages_raw_train": (
+            "odinw/ShellfishOpenImages/raw/train/",
+            "odinw/ShellfishOpenImages/raw/train/annotations_without_background_converted.json",
+        ),
+        "odinw_ShellfishOpenImages_raw_val": (
+            "odinw/ShellfishOpenImages/raw/valid/",
+            "odinw/ShellfishOpenImages/raw/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_ShellfishOpenImages_raw_test": (
+            "odinw/ShellfishOpenImages/raw/test/",
+            "odinw/ShellfishOpenImages/raw/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_ThermalCheetah": {
+        "odinw_ThermalCheetah_train": (
+            "odinw/ThermalCheetah/train/",
+            "odinw/ThermalCheetah/train/annotations_without_background_converted.json",
+        ),
+        "odinw_ThermalCheetah_val": (
+            "odinw/ThermalCheetah/valid/",
+            "odinw/ThermalCheetah/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_ThermalCheetah_test": (
+            "odinw/ThermalCheetah/test/",
+            "odinw/ThermalCheetah/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_thermalDogsAndPeople": {
+        "odinw_thermalDogsAndPeople_train": (
+            "odinw/thermalDogsAndPeople/train/",
+            "odinw/thermalDogsAndPeople/train/annotations_without_background_converted.json",
+        ),
+        "odinw_thermalDogsAndPeople_val": (
+            "odinw/thermalDogsAndPeople/valid/",
+            "odinw/thermalDogsAndPeople/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_thermalDogsAndPeople_test": (
+            "odinw/thermalDogsAndPeople/test/",
+            "odinw/thermalDogsAndPeople/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_UnoCards_raw": {
+        "odinw_UnoCards_raw_train": (
+            "odinw/UnoCards/raw/train/",
+            "odinw/UnoCards/raw/train/annotations_without_background_converted.json",
+        ),
+        "odinw_UnoCards_raw_val": (
+            "odinw/UnoCards/raw/valid/",
+            "odinw/UnoCards/raw/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_UnoCards_raw_minival": (
+            "odinw/UnoCards/raw/mini_val/",
+            "odinw/UnoCards/raw/mini_val/annotations_without_background_converted.json",
+        ),
+        "odinw_UnoCards_raw_test": (
+            "odinw/UnoCards/raw/test/",
+            "odinw/UnoCards/raw/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_VehiclesOpenImages_416x416": {
+        "odinw_VehiclesOpenImages_416x416_train": (
+            "odinw/VehiclesOpenImages/416x416/train/",
+            "odinw/VehiclesOpenImages/416x416/train/annotations_without_background_converted.json",
+        ),
+        "odinw_VehiclesOpenImages_416x416_val": (
+            "odinw/VehiclesOpenImages/416x416/valid/",
+            "odinw/VehiclesOpenImages/416x416/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_VehiclesOpenImages_416x416_minival": (
+            "odinw/VehiclesOpenImages/416x416/mini_val/",
+            "odinw/VehiclesOpenImages/416x416/mini_val/annotations_without_background_converted.json",
+        ),
+        "odinw_VehiclesOpenImages_416x416_test": (
+            "odinw/VehiclesOpenImages/416x416/test/",
+            "odinw/VehiclesOpenImages/416x416/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_websiteScreenshots": {
+        "odinw_websiteScreenshots_train": (
+            "odinw/websiteScreenshots/train/",
+            "odinw/websiteScreenshots/train/annotations_without_background_converted.json",
+        ),
+        "odinw_websiteScreenshots_val": (
+            "odinw/websiteScreenshots/valid/",
+            "odinw/websiteScreenshots/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_websiteScreenshots_minival": (
+            "odinw/websiteScreenshots/mini_val/",
+            "odinw/websiteScreenshots/mini_val/annotations_without_background_converted.json",
+        ),
+        "odinw_websiteScreenshots_test": (
+            "odinw/websiteScreenshots/test/",
+            "odinw/websiteScreenshots/test/annotations_without_background_converted.json",
+        ),
+    },
+    "odinw_WildfireSmoke": {
+        "odinw_WildfireSmoke_train": (
+            "odinw/WildfireSmoke/train/",
+            "odinw/WildfireSmoke/train/annotations_without_background_converted.json",
+        ),
+        "odinw_WildfireSmoke_val": (
+            "odinw/WildfireSmoke/valid/",
+            "odinw/WildfireSmoke/valid/annotations_without_background_converted.json",
+        ),
+        "odinw_WildfireSmoke_test": (
+            "odinw/WildfireSmoke/test/",
+            "odinw/WildfireSmoke/test/annotations_without_background_converted.json",
+        ),
+    },
+}
+
+
+def _get_builtin_metadata(name):
+    meta = {}
+    if name.split("_")[1] in ODINW_PROMPTS:
+        meta["thing_classes"] = [
+            ODINW_PROMPTS[name.split("_")[1]](m["name"])
+            for m in ODINW_CATEGORIES[name.split("_")[1]]
+        ]
+    else:
+        meta["thing_classes"] = [m["name"] for m in ODINW_CATEGORIES[name.split("_")[1]]]
+    return meta
+
+
+def register_all_odinw(root):
+    for dataset_name, splits_per_dataset in _PREDEFINED_SPLITS_ODINW.items():
+        for key, (image_root, json_file) in splits_per_dataset.items():
+            register_coco_instances(
+                key,
+                _get_builtin_metadata(dataset_name),
+                os.path.join(root, json_file) if "://" not in json_file else json_file,
+                os.path.join(root, image_root),
+            )
+
+
+if __name__.endswith(".odinw_instance"):
+    # Assume pre-defined datasets live in `./datasets`.
+    _root = os.getenv("DATASET", "datasets")
+    register_all_odinw(_root)

approach/ovod/APE/ape/data/datasets/odinw_prompts.py

@@ -0,0 +1,75 @@
+ODINW_PROMPTS = {
+    "AerialMaritimeDrone": lambda name: "a ship" if name == "boat" else name,
+    "AmericanSignLanguageLetters": lambda name: "hand gesture '{}'".format(name),
+    "BCCD": lambda name: "Red-Blood-Cell-(RBC)"
+    if name == "RBC"
+    else "White-Blood-Cell-(WBC)"
+    if name == "WBC"
+    else "Blood-Platelet-Cell-(BPC)"
+    if name == "Platelets"
+    else name,
+    "boggleBoards": lambda name: "letter '{}'".format(name.upper()),
+    "brackishUnderwater": lambda name: "big_fish" if name == "fish" else name,
+    "ChessPieces": lambda name: "chess piece {}".format(name),
+    "dice": lambda name: "dice {}".format(name),
+    "DroneControl": lambda name: "body gesture '{}'".format(name),
+    "EgoHands-specific": lambda name: "{} hand".format(name),
+    # "EgoHands-specific": lambda name: "my left hand"
+    # if name == "myleft"
+    # else "my right hand"
+    # if name == "myright"
+    # else "your right hand"
+    # if name == "yourright"
+    # else "your left hand"
+    # if name == "yourleft"
+    # else name,
+    "HardHatWorkers": lambda name: "human head wearing a helmet"
+    if name == "helmet"
+    else "human head"
+    if name == "head"
+    else name,
+    "MaskWearing": lambda name: "human head wearing a mask"
+    if name == "mask"
+    else "human head"
+    if name == "no-mask"
+    else name,
+    "MountainDewCommercial": lambda name: "small {}".format(name),
+    "NorthAmericaMushrooms": lambda name: "mushroom {}".format(name),
+    "openPoetryVision": lambda name: "some text with font {}".format(name),
+    "OxfordPets-by-breed": lambda name: "head of {}".format(name),
+    "OxfordPets-by-species": lambda name: "head of {}".format(name),
+    "PKLot": lambda name: "{} parking slot".format(name),
+    "pothole": lambda name: "broken {}".format(name),
+    "ThermalCheetah": lambda name: "person" if name == "human" else name,
+    "UnoCards": lambda name: "Arabic numerals 0"
+    if name == "0"
+    else "Arabic numerals 1"
+    if name == "1"
+    else "Arabic numerals +4"
+    if name == "2"
+    else "Arabic numerals +2"
+    if name == "3"
+    else "two arrows"
+    if name == "4"
+    else "cross cycle"
+    if name == "5"
+    else "colorful cycle"
+    if name == "6"
+    else "Arabic numerals 2"
+    if name == "7"
+    else "Arabic numerals 3"
+    if name == "8"
+    else "Arabic numerals 4"
+    if name == "9"
+    else "Arabic numerals 5"
+    if name == "10"
+    else "Arabic numerals 6"
+    if name == "11"
+    else "Arabic numerals 7"
+    if name == "12"
+    else "Arabic numerals 8"
+    if name == "13"
+    else "Arabic numerals 9"
+    if name == "14"
+    else name,
+}

approach/ovod/APE/ape/data/datasets/oid.py

@@ -0,0 +1,1573 @@
+import contextlib
+import io
+import logging
+import os
+
+from detectron2.data import DatasetCatalog, MetadataCatalog
+
+from .coco import custom_load_coco_json
+from .openimages_v6_category_image_count import OPENIMAGES_v6_CATEGORY_IMAGE_COUNT
+
+
+def register_oid_instances(name, metadata, json_file, image_root):
+    """
+    Register a dataset in COCO's json annotation format for
+    instance detection, instance segmentation and keypoint detection.
+    (i.e., Type 1 and 2 in http://cocodataset.org/#format-data.
+    `instances*.json` and `person_keypoints*.json` in the dataset).
+
+    This is an example of how to register a new dataset.
+    You can do something similar to this function, to register new datasets.
+
+    Args:
+        name (str): the name that identifies a dataset, e.g. "coco_2014_train".
+        metadata (dict): extra metadata associated with this dataset.  You can
+            leave it as an empty dict.
+        json_file (str): path to the json instance annotation file.
+        image_root (str or path-like): directory which contains all the images.
+    """
+    assert isinstance(name, str), name
+    assert isinstance(json_file, (str, os.PathLike)), json_file
+    assert isinstance(image_root, (str, os.PathLike)), image_root
+    # 1. register a function which returns dicts
+    DatasetCatalog.register(name, lambda: custom_load_coco_json(json_file, image_root, name))
+
+    # 2. Optionally, add metadata about this dataset,
+    # since they might be useful in evaluation, visualization or logging
+    MetadataCatalog.get(name).set(
+        json_file=json_file, image_root=image_root, evaluator_type="oid", **metadata
+    )
+
+
+OPENIMAGES_2019_CATEGORIES = [
+    {"id": 1, "name": "Infant bed", "freebase_id": "/m/061hd_"},
+    {"id": 2, "name": "Rose", "freebase_id": "/m/06m11"},
+    {"id": 3, "name": "Flag", "freebase_id": "/m/03120"},
+    {"id": 4, "name": "Flashlight", "freebase_id": "/m/01kb5b"},
+    {"id": 5, "name": "Sea turtle", "freebase_id": "/m/0120dh"},
+    {"id": 6, "name": "Camera", "freebase_id": "/m/0dv5r"},
+    {"id": 7, "name": "Animal", "freebase_id": "/m/0jbk"},
+    {"id": 8, "name": "Glove", "freebase_id": "/m/0174n1"},
+    {"id": 9, "name": "Crocodile", "freebase_id": "/m/09f_2"},
+    {"id": 10, "name": "Cattle", "freebase_id": "/m/01xq0k1"},
+    {"id": 11, "name": "House", "freebase_id": "/m/03jm5"},
+    {"id": 12, "name": "Guacamole", "freebase_id": "/m/02g30s"},
+    {"id": 13, "name": "Penguin", "freebase_id": "/m/05z6w"},
+    {"id": 14, "name": "Vehicle registration plate", "freebase_id": "/m/01jfm_"},
+    {"id": 15, "name": "Bench", "freebase_id": "/m/076lb9"},
+    {"id": 16, "name": "Ladybug", "freebase_id": "/m/0gj37"},
+    {"id": 17, "name": "Human nose", "freebase_id": "/m/0k0pj"},
+    {"id": 18, "name": "Watermelon", "freebase_id": "/m/0kpqd"},
+    {"id": 19, "name": "Flute", "freebase_id": "/m/0l14j_"},
+    {"id": 20, "name": "Butterfly", "freebase_id": "/m/0cyf8"},
+    {"id": 21, "name": "Washing machine", "freebase_id": "/m/0174k2"},
+    {"id": 22, "name": "Raccoon", "freebase_id": "/m/0dq75"},
+    {"id": 23, "name": "Segway", "freebase_id": "/m/076bq"},
+    {"id": 24, "name": "Taco", "freebase_id": "/m/07crc"},
+    {"id": 25, "name": "Jellyfish", "freebase_id": "/m/0d8zb"},
+    {"id": 26, "name": "Cake", "freebase_id": "/m/0fszt"},
+    {"id": 27, "name": "Pen", "freebase_id": "/m/0k1tl"},
+    {"id": 28, "name": "Cannon", "freebase_id": "/m/020kz"},
+    {"id": 29, "name": "Bread", "freebase_id": "/m/09728"},
+    {"id": 30, "name": "Tree", "freebase_id": "/m/07j7r"},
+    {"id": 31, "name": "Shellfish", "freebase_id": "/m/0fbdv"},
+    {"id": 32, "name": "Bed", "freebase_id": "/m/03ssj5"},
+    {"id": 33, "name": "Hamster", "freebase_id": "/m/03qrc"},
+    {"id": 34, "name": "Hat", "freebase_id": "/m/02dl1y"},
+    {"id": 35, "name": "Toaster", "freebase_id": "/m/01k6s3"},
+    {"id": 36, "name": "Sombrero", "freebase_id": "/m/02jfl0"},
+    {"id": 37, "name": "Tiara", "freebase_id": "/m/01krhy"},
+    {"id": 38, "name": "Bowl", "freebase_id": "/m/04kkgm"},
+    {"id": 39, "name": "Dragonfly", "freebase_id": "/m/0ft9s"},
+    {"id": 40, "name": "Moths and butterflies", "freebase_id": "/m/0d_2m"},
+    {"id": 41, "name": "Antelope", "freebase_id": "/m/0czz2"},
+    {"id": 42, "name": "Vegetable", "freebase_id": "/m/0f4s2w"},
+    {"id": 43, "name": "Torch", "freebase_id": "/m/07dd4"},
+    {"id": 44, "name": "Building", "freebase_id": "/m/0cgh4"},
+    {"id": 45, "name": "Power plugs and sockets", "freebase_id": "/m/03bbps"},
+    {"id": 46, "name": "Blender", "freebase_id": "/m/02pjr4"},
+    {"id": 47, "name": "Billiard table", "freebase_id": "/m/04p0qw"},
+    {"id": 48, "name": "Cutting board", "freebase_id": "/m/02pdsw"},
+    {"id": 49, "name": "Bronze sculpture", "freebase_id": "/m/01yx86"},
+    {"id": 50, "name": "Turtle", "freebase_id": "/m/09dzg"},
+    {"id": 51, "name": "Broccoli", "freebase_id": "/m/0hkxq"},
+    {"id": 52, "name": "Tiger", "freebase_id": "/m/07dm6"},
+    {"id": 53, "name": "Mirror", "freebase_id": "/m/054_l"},
+    {"id": 54, "name": "Bear", "freebase_id": "/m/01dws"},
+    {"id": 55, "name": "Zucchini", "freebase_id": "/m/027pcv"},
+    {"id": 56, "name": "Dress", "freebase_id": "/m/01d40f"},
+    {"id": 57, "name": "Volleyball", "freebase_id": "/m/02rgn06"},
+    {"id": 58, "name": "Guitar", "freebase_id": "/m/0342h"},
+    {"id": 59, "name": "Reptile", "freebase_id": "/m/06bt6"},
+    {"id": 60, "name": "Golf cart", "freebase_id": "/m/0323sq"},
+    {"id": 61, "name": "Tart", "freebase_id": "/m/02zvsm"},
+    {"id": 62, "name": "Fedora", "freebase_id": "/m/02fq_6"},
+    {"id": 63, "name": "Carnivore", "freebase_id": "/m/01lrl"},
+    {"id": 64, "name": "Car", "freebase_id": "/m/0k4j"},
+    {"id": 65, "name": "Lighthouse", "freebase_id": "/m/04h7h"},
+    {"id": 66, "name": "Coffeemaker", "freebase_id": "/m/07xyvk"},
+    {"id": 67, "name": "Food processor", "freebase_id": "/m/03y6mg"},
+    {"id": 68, "name": "Truck", "freebase_id": "/m/07r04"},
+    {"id": 69, "name": "Bookcase", "freebase_id": "/m/03__z0"},
+    {"id": 70, "name": "Surfboard", "freebase_id": "/m/019w40"},
+    {"id": 71, "name": "Footwear", "freebase_id": "/m/09j5n"},
+    {"id": 72, "name": "Bench", "freebase_id": "/m/0cvnqh"},
+    {"id": 73, "name": "Necklace", "freebase_id": "/m/01llwg"},
+    {"id": 74, "name": "Flower", "freebase_id": "/m/0c9ph5"},
+    {"id": 75, "name": "Radish", "freebase_id": "/m/015x5n"},
+    {"id": 76, "name": "Marine mammal", "freebase_id": "/m/0gd2v"},
+    {"id": 77, "name": "Frying pan", "freebase_id": "/m/04v6l4"},
+    {"id": 78, "name": "Tap", "freebase_id": "/m/02jz0l"},
+    {"id": 79, "name": "Peach", "freebase_id": "/m/0dj6p"},
+    {"id": 80, "name": "Knife", "freebase_id": "/m/04ctx"},
+    {"id": 81, "name": "Handbag", "freebase_id": "/m/080hkjn"},
+    {"id": 82, "name": "Laptop", "freebase_id": "/m/01c648"},
+    {"id": 83, "name": "Tent", "freebase_id": "/m/01j61q"},
+    {"id": 84, "name": "Ambulance", "freebase_id": "/m/012n7d"},
+    {"id": 85, "name": "Christmas tree", "freebase_id": "/m/025nd"},
+    {"id": 86, "name": "Eagle", "freebase_id": "/m/09csl"},
+    {"id": 87, "name": "Limousine", "freebase_id": "/m/01lcw4"},
+    {"id": 88, "name": "Kitchen & dining room table", "freebase_id": "/m/0h8n5zk"},
+    {"id": 89, "name": "Polar bear", "freebase_id": "/m/0633h"},
+    {"id": 90, "name": "Tower", "freebase_id": "/m/01fdzj"},
+    {"id": 91, "name": "Football", "freebase_id": "/m/01226z"},
+    {"id": 92, "name": "Willow", "freebase_id": "/m/0mw_6"},
+    {"id": 93, "name": "Human head", "freebase_id": "/m/04hgtk"},
+    {"id": 94, "name": "Stop sign", "freebase_id": "/m/02pv19"},
+    {"id": 95, "name": "Banana", "freebase_id": "/m/09qck"},
+    {"id": 96, "name": "Mixer", "freebase_id": "/m/063rgb"},
+    {"id": 97, "name": "Binoculars", "freebase_id": "/m/0lt4_"},
+    {"id": 98, "name": "Dessert", "freebase_id": "/m/0270h"},
+    {"id": 99, "name": "Bee", "freebase_id": "/m/01h3n"},
+    {"id": 100, "name": "Chair", "freebase_id": "/m/01mzpv"},
+    {"id": 101, "name": "Wood-burning stove", "freebase_id": "/m/04169hn"},
+    {"id": 102, "name": "Flowerpot", "freebase_id": "/m/0fm3zh"},
+    {"id": 103, "name": "Beaker", "freebase_id": "/m/0d20w4"},
+    {"id": 104, "name": "Oyster", "freebase_id": "/m/0_cp5"},
+    {"id": 105, "name": "Woodpecker", "freebase_id": "/m/01dy8n"},
+    {"id": 106, "name": "Harp", "freebase_id": "/m/03m5k"},
+    {"id": 107, "name": "Bathtub", "freebase_id": "/m/03dnzn"},
+    {"id": 108, "name": "Wall clock", "freebase_id": "/m/0h8mzrc"},
+    {"id": 109, "name": "Sports uniform", "freebase_id": "/m/0h8mhzd"},
+    {"id": 110, "name": "Rhinoceros", "freebase_id": "/m/03d443"},
+    {"id": 111, "name": "Beehive", "freebase_id": "/m/01gllr"},
+    {"id": 112, "name": "Cupboard", "freebase_id": "/m/0642b4"},
+    {"id": 113, "name": "Chicken", "freebase_id": "/m/09b5t"},
+    {"id": 114, "name": "Man", "freebase_id": "/m/04yx4"},
+    {"id": 115, "name": "Blue jay", "freebase_id": "/m/01f8m5"},
+    {"id": 116, "name": "Cucumber", "freebase_id": "/m/015x4r"},
+    {"id": 117, "name": "Balloon", "freebase_id": "/m/01j51"},
+    {"id": 118, "name": "Kite", "freebase_id": "/m/02zt3"},
+    {"id": 119, "name": "Fireplace", "freebase_id": "/m/03tw93"},
+    {"id": 120, "name": "Lantern", "freebase_id": "/m/01jfsr"},
+    {"id": 121, "name": "Missile", "freebase_id": "/m/04ylt"},
+    {"id": 122, "name": "Book", "freebase_id": "/m/0bt_c3"},
+    {"id": 123, "name": "Spoon", "freebase_id": "/m/0cmx8"},
+    {"id": 124, "name": "Grapefruit", "freebase_id": "/m/0hqkz"},
+    {"id": 125, "name": "Squirrel", "freebase_id": "/m/071qp"},
+    {"id": 126, "name": "Orange", "freebase_id": "/m/0cyhj_"},
+    {"id": 127, "name": "Coat", "freebase_id": "/m/01xygc"},
+    {"id": 128, "name": "Punching bag", "freebase_id": "/m/0420v5"},
+    {"id": 129, "name": "Zebra", "freebase_id": "/m/0898b"},
+    {"id": 130, "name": "Billboard", "freebase_id": "/m/01knjb"},
+    {"id": 131, "name": "Bicycle", "freebase_id": "/m/0199g"},
+    {"id": 132, "name": "Door handle", "freebase_id": "/m/03c7gz"},
+    {"id": 133, "name": "Mechanical fan", "freebase_id": "/m/02x984l"},
+    {"id": 134, "name": "Ring binder", "freebase_id": "/m/04zwwv"},
+    {"id": 135, "name": "Table", "freebase_id": "/m/04bcr3"},
+    {"id": 136, "name": "Parrot", "freebase_id": "/m/0gv1x"},
+    {"id": 137, "name": "Sock", "freebase_id": "/m/01nq26"},
+    {"id": 138, "name": "Vase", "freebase_id": "/m/02s195"},
+    {"id": 139, "name": "Weapon", "freebase_id": "/m/083kb"},
+    {"id": 140, "name": "Shotgun", "freebase_id": "/m/06nrc"},
+    {"id": 141, "name": "Glasses", "freebase_id": "/m/0jyfg"},
+    {"id": 142, "name": "Seahorse", "freebase_id": "/m/0nybt"},
+    {"id": 143, "name": "Belt", "freebase_id": "/m/0176mf"},
+    {"id": 144, "name": "Watercraft", "freebase_id": "/m/01rzcn"},
+    {"id": 145, "name": "Window", "freebase_id": "/m/0d4v4"},
+    {"id": 146, "name": "Giraffe", "freebase_id": "/m/03bk1"},
+    {"id": 147, "name": "Lion", "freebase_id": "/m/096mb"},
+    {"id": 148, "name": "Tire", "freebase_id": "/m/0h9mv"},
+    {"id": 149, "name": "Vehicle", "freebase_id": "/m/07yv9"},
+    {"id": 150, "name": "Canoe", "freebase_id": "/m/0ph39"},
+    {"id": 151, "name": "Tie", "freebase_id": "/m/01rkbr"},
+    {"id": 152, "name": "Shelf", "freebase_id": "/m/0gjbg72"},
+    {"id": 153, "name": "Picture frame", "freebase_id": "/m/06z37_"},
+    {"id": 154, "name": "Printer", "freebase_id": "/m/01m4t"},
+    {"id": 155, "name": "Human leg", "freebase_id": "/m/035r7c"},
+    {"id": 156, "name": "Boat", "freebase_id": "/m/019jd"},
+    {"id": 157, "name": "Slow cooker", "freebase_id": "/m/02tsc9"},
+    {"id": 158, "name": "Croissant", "freebase_id": "/m/015wgc"},
+    {"id": 159, "name": "Candle", "freebase_id": "/m/0c06p"},
+    {"id": 160, "name": "Pancake", "freebase_id": "/m/01dwwc"},
+    {"id": 161, "name": "Pillow", "freebase_id": "/m/034c16"},
+    {"id": 162, "name": "Coin", "freebase_id": "/m/0242l"},
+    {"id": 163, "name": "Stretcher", "freebase_id": "/m/02lbcq"},
+    {"id": 164, "name": "Sandal", "freebase_id": "/m/03nfch"},
+    {"id": 165, "name": "Woman", "freebase_id": "/m/03bt1vf"},
+    {"id": 166, "name": "Stairs", "freebase_id": "/m/01lynh"},
+    {"id": 167, "name": "Harpsichord", "freebase_id": "/m/03q5t"},
+    {"id": 168, "name": "Stool", "freebase_id": "/m/0fqt361"},
+    {"id": 169, "name": "Bus", "freebase_id": "/m/01bjv"},
+    {"id": 170, "name": "Suitcase", "freebase_id": "/m/01s55n"},
+    {"id": 171, "name": "Human mouth", "freebase_id": "/m/0283dt1"},
+    {"id": 172, "name": "Juice", "freebase_id": "/m/01z1kdw"},
+    {"id": 173, "name": "Skull", "freebase_id": "/m/016m2d"},
+    {"id": 174, "name": "Door", "freebase_id": "/m/02dgv"},
+    {"id": 175, "name": "Violin", "freebase_id": "/m/07y_7"},
+    {"id": 176, "name": "Chopsticks", "freebase_id": "/m/01_5g"},
+    {"id": 177, "name": "Digital clock", "freebase_id": "/m/06_72j"},
+    {"id": 178, "name": "Sunflower", "freebase_id": "/m/0ftb8"},
+    {"id": 179, "name": "Leopard", "freebase_id": "/m/0c29q"},
+    {"id": 180, "name": "Bell pepper", "freebase_id": "/m/0jg57"},
+    {"id": 181, "name": "Harbor seal", "freebase_id": "/m/02l8p9"},
+    {"id": 182, "name": "Snake", "freebase_id": "/m/078jl"},
+    {"id": 183, "name": "Sewing machine", "freebase_id": "/m/0llzx"},
+    {"id": 184, "name": "Goose", "freebase_id": "/m/0dbvp"},
+    {"id": 185, "name": "Helicopter", "freebase_id": "/m/09ct_"},
+    {"id": 186, "name": "Seat belt", "freebase_id": "/m/0dkzw"},
+    {"id": 187, "name": "Coffee cup", "freebase_id": "/m/02p5f1q"},
+    {"id": 188, "name": "Microwave oven", "freebase_id": "/m/0fx9l"},
+    {"id": 189, "name": "Hot dog", "freebase_id": "/m/01b9xk"},
+    {"id": 190, "name": "Countertop", "freebase_id": "/m/0b3fp9"},
+    {"id": 191, "name": "Serving tray", "freebase_id": "/m/0h8n27j"},
+    {"id": 192, "name": "Dog bed", "freebase_id": "/m/0h8n6f9"},
+    {"id": 193, "name": "Beer", "freebase_id": "/m/01599"},
+    {"id": 194, "name": "Sunglasses", "freebase_id": "/m/017ftj"},
+    {"id": 195, "name": "Golf ball", "freebase_id": "/m/044r5d"},
+    {"id": 196, "name": "Waffle", "freebase_id": "/m/01dwsz"},
+    {"id": 197, "name": "Palm tree", "freebase_id": "/m/0cdl1"},
+    {"id": 198, "name": "Trumpet", "freebase_id": "/m/07gql"},
+    {"id": 199, "name": "Ruler", "freebase_id": "/m/0hdln"},
+    {"id": 200, "name": "Helmet", "freebase_id": "/m/0zvk5"},
+    {"id": 201, "name": "Ladder", "freebase_id": "/m/012w5l"},
+    {"id": 202, "name": "Office building", "freebase_id": "/m/021sj1"},
+    {"id": 203, "name": "Tablet computer", "freebase_id": "/m/0bh9flk"},
+    {"id": 204, "name": "Toilet paper", "freebase_id": "/m/09gtd"},
+    {"id": 205, "name": "Pomegranate", "freebase_id": "/m/0jwn_"},
+    {"id": 206, "name": "Skirt", "freebase_id": "/m/02wv6h6"},
+    {"id": 207, "name": "Gas stove", "freebase_id": "/m/02wv84t"},
+    {"id": 208, "name": "Cookie", "freebase_id": "/m/021mn"},
+    {"id": 209, "name": "Cart", "freebase_id": "/m/018p4k"},
+    {"id": 210, "name": "Raven", "freebase_id": "/m/06j2d"},
+    {"id": 211, "name": "Egg", "freebase_id": "/m/033cnk"},
+    {"id": 212, "name": "Burrito", "freebase_id": "/m/01j3zr"},
+    {"id": 213, "name": "Goat", "freebase_id": "/m/03fwl"},
+    {"id": 214, "name": "Kitchen knife", "freebase_id": "/m/058qzx"},
+    {"id": 215, "name": "Skateboard", "freebase_id": "/m/06_fw"},
+    {"id": 216, "name": "Salt and pepper shakers", "freebase_id": "/m/02x8cch"},
+    {"id": 217, "name": "Lynx", "freebase_id": "/m/04g2r"},
+    {"id": 218, "name": "Boot", "freebase_id": "/m/01b638"},
+    {"id": 219, "name": "Platter", "freebase_id": "/m/099ssp"},
+    {"id": 220, "name": "Ski", "freebase_id": "/m/071p9"},
+    {"id": 221, "name": "Swimwear", "freebase_id": "/m/01gkx_"},
+    {"id": 222, "name": "Swimming pool", "freebase_id": "/m/0b_rs"},
+    {"id": 223, "name": "Drinking straw", "freebase_id": "/m/03v5tg"},
+    {"id": 224, "name": "Wrench", "freebase_id": "/m/01j5ks"},
+    {"id": 225, "name": "Drum", "freebase_id": "/m/026t6"},
+    {"id": 226, "name": "Ant", "freebase_id": "/m/0_k2"},
+    {"id": 227, "name": "Human ear", "freebase_id": "/m/039xj_"},
+    {"id": 228, "name": "Headphones", "freebase_id": "/m/01b7fy"},
+    {"id": 229, "name": "Fountain", "freebase_id": "/m/0220r2"},
+    {"id": 230, "name": "Bird", "freebase_id": "/m/015p6"},
+    {"id": 231, "name": "Jeans", "freebase_id": "/m/0fly7"},
+    {"id": 232, "name": "Television", "freebase_id": "/m/07c52"},
+    {"id": 233, "name": "Crab", "freebase_id": "/m/0n28_"},
+    {"id": 234, "name": "Microphone", "freebase_id": "/m/0hg7b"},
+    {"id": 235, "name": "Home appliance", "freebase_id": "/m/019dx1"},
+    {"id": 236, "name": "Snowplow", "freebase_id": "/m/04vv5k"},
+    {"id": 237, "name": "Beetle", "freebase_id": "/m/020jm"},
+    {"id": 238, "name": "Artichoke", "freebase_id": "/m/047v4b"},
+    {"id": 239, "name": "Jet ski", "freebase_id": "/m/01xs3r"},
+    {"id": 240, "name": "Stationary bicycle", "freebase_id": "/m/03kt2w"},
+    {"id": 241, "name": "Human hair", "freebase_id": "/m/03q69"},
+    {"id": 242, "name": "Brown bear", "freebase_id": "/m/01dxs"},
+    {"id": 243, "name": "Starfish", "freebase_id": "/m/01h8tj"},
+    {"id": 244, "name": "Fork", "freebase_id": "/m/0dt3t"},
+    {"id": 245, "name": "Lobster", "freebase_id": "/m/0cjq5"},
+    {"id": 246, "name": "Corded phone", "freebase_id": "/m/0h8lkj8"},
+    {"id": 247, "name": "Drink", "freebase_id": "/m/0271t"},
+    {"id": 248, "name": "Saucer", "freebase_id": "/m/03q5c7"},
+    {"id": 249, "name": "Carrot", "freebase_id": "/m/0fj52s"},
+    {"id": 250, "name": "Insect", "freebase_id": "/m/03vt0"},
+    {"id": 251, "name": "Clock", "freebase_id": "/m/01x3z"},
+    {"id": 252, "name": "Castle", "freebase_id": "/m/0d5gx"},
+    {"id": 253, "name": "Tennis racket", "freebase_id": "/m/0h8my_4"},
+    {"id": 254, "name": "Ceiling fan", "freebase_id": "/m/03ldnb"},
+    {"id": 255, "name": "Asparagus", "freebase_id": "/m/0cjs7"},
+    {"id": 256, "name": "Jaguar", "freebase_id": "/m/0449p"},
+    {"id": 257, "name": "Musical instrument", "freebase_id": "/m/04szw"},
+    {"id": 258, "name": "Train", "freebase_id": "/m/07jdr"},
+    {"id": 259, "name": "Cat", "freebase_id": "/m/01yrx"},
+    {"id": 260, "name": "Rifle", "freebase_id": "/m/06c54"},
+    {"id": 261, "name": "Dumbbell", "freebase_id": "/m/04h8sr"},
+    {"id": 262, "name": "Mobile phone", "freebase_id": "/m/050k8"},
+    {"id": 263, "name": "Taxi", "freebase_id": "/m/0pg52"},
+    {"id": 264, "name": "Shower", "freebase_id": "/m/02f9f_"},
+    {"id": 265, "name": "Pitcher", "freebase_id": "/m/054fyh"},
+    {"id": 266, "name": "Lemon", "freebase_id": "/m/09k_b"},
+    {"id": 267, "name": "Invertebrate", "freebase_id": "/m/03xxp"},
+    {"id": 268, "name": "Turkey", "freebase_id": "/m/0jly1"},
+    {"id": 269, "name": "High heels", "freebase_id": "/m/06k2mb"},
+    {"id": 270, "name": "Bust", "freebase_id": "/m/04yqq2"},
+    {"id": 271, "name": "Elephant", "freebase_id": "/m/0bwd_0j"},
+    {"id": 272, "name": "Scarf", "freebase_id": "/m/02h19r"},
+    {"id": 273, "name": "Barrel", "freebase_id": "/m/02zn6n"},
+    {"id": 274, "name": "Trombone", "freebase_id": "/m/07c6l"},
+    {"id": 275, "name": "Pumpkin", "freebase_id": "/m/05zsy"},
+    {"id": 276, "name": "Box", "freebase_id": "/m/025dyy"},
+    {"id": 277, "name": "Tomato", "freebase_id": "/m/07j87"},
+    {"id": 278, "name": "Frog", "freebase_id": "/m/09ld4"},
+    {"id": 279, "name": "Bidet", "freebase_id": "/m/01vbnl"},
+    {"id": 280, "name": "Human face", "freebase_id": "/m/0dzct"},
+    {"id": 281, "name": "Houseplant", "freebase_id": "/m/03fp41"},
+    {"id": 282, "name": "Van", "freebase_id": "/m/0h2r6"},
+    {"id": 283, "name": "Shark", "freebase_id": "/m/0by6g"},
+    {"id": 284, "name": "Ice cream", "freebase_id": "/m/0cxn2"},
+    {"id": 285, "name": "Swim cap", "freebase_id": "/m/04tn4x"},
+    {"id": 286, "name": "Falcon", "freebase_id": "/m/0f6wt"},
+    {"id": 287, "name": "Ostrich", "freebase_id": "/m/05n4y"},
+    {"id": 288, "name": "Handgun", "freebase_id": "/m/0gxl3"},
+    {"id": 289, "name": "Whiteboard", "freebase_id": "/m/02d9qx"},
+    {"id": 290, "name": "Lizard", "freebase_id": "/m/04m9y"},
+    {"id": 291, "name": "Pasta", "freebase_id": "/m/05z55"},
+    {"id": 292, "name": "Snowmobile", "freebase_id": "/m/01x3jk"},
+    {"id": 293, "name": "Light bulb", "freebase_id": "/m/0h8l4fh"},
+    {"id": 294, "name": "Window blind", "freebase_id": "/m/031b6r"},
+    {"id": 295, "name": "Muffin", "freebase_id": "/m/01tcjp"},
+    {"id": 296, "name": "Pretzel", "freebase_id": "/m/01f91_"},
+    {"id": 297, "name": "Computer monitor", "freebase_id": "/m/02522"},
+    {"id": 298, "name": "Horn", "freebase_id": "/m/0319l"},
+    {"id": 299, "name": "Furniture", "freebase_id": "/m/0c_jw"},
+    {"id": 300, "name": "Sandwich", "freebase_id": "/m/0l515"},
+    {"id": 301, "name": "Fox", "freebase_id": "/m/0306r"},
+    {"id": 302, "name": "Convenience store", "freebase_id": "/m/0crjs"},
+    {"id": 303, "name": "Fish", "freebase_id": "/m/0ch_cf"},
+    {"id": 304, "name": "Fruit", "freebase_id": "/m/02xwb"},
+    {"id": 305, "name": "Earrings", "freebase_id": "/m/01r546"},
+    {"id": 306, "name": "Curtain", "freebase_id": "/m/03rszm"},
+    {"id": 307, "name": "Grape", "freebase_id": "/m/0388q"},
+    {"id": 308, "name": "Sofa bed", "freebase_id": "/m/03m3pdh"},
+    {"id": 309, "name": "Horse", "freebase_id": "/m/03k3r"},
+    {"id": 310, "name": "Luggage and bags", "freebase_id": "/m/0hf58v5"},
+    {"id": 311, "name": "Desk", "freebase_id": "/m/01y9k5"},
+    {"id": 312, "name": "Crutch", "freebase_id": "/m/05441v"},
+    {"id": 313, "name": "Bicycle helmet", "freebase_id": "/m/03p3bw"},
+    {"id": 314, "name": "Tick", "freebase_id": "/m/0175cv"},
+    {"id": 315, "name": "Airplane", "freebase_id": "/m/0cmf2"},
+    {"id": 316, "name": "Canary", "freebase_id": "/m/0ccs93"},
+    {"id": 317, "name": "Spatula", "freebase_id": "/m/02d1br"},
+    {"id": 318, "name": "Watch", "freebase_id": "/m/0gjkl"},
+    {"id": 319, "name": "Lily", "freebase_id": "/m/0jqgx"},
+    {"id": 320, "name": "Kitchen appliance", "freebase_id": "/m/0h99cwc"},
+    {"id": 321, "name": "Filing cabinet", "freebase_id": "/m/047j0r"},
+    {"id": 322, "name": "Aircraft", "freebase_id": "/m/0k5j"},
+    {"id": 323, "name": "Cake stand", "freebase_id": "/m/0h8n6ft"},
+    {"id": 324, "name": "Candy", "freebase_id": "/m/0gm28"},
+    {"id": 325, "name": "Sink", "freebase_id": "/m/0130jx"},
+    {"id": 326, "name": "Mouse", "freebase_id": "/m/04rmv"},
+    {"id": 327, "name": "Wine", "freebase_id": "/m/081qc"},
+    {"id": 328, "name": "Wheelchair", "freebase_id": "/m/0qmmr"},
+    {"id": 329, "name": "Goldfish", "freebase_id": "/m/03fj2"},
+    {"id": 330, "name": "Refrigerator", "freebase_id": "/m/040b_t"},
+    {"id": 331, "name": "French fries", "freebase_id": "/m/02y6n"},
+    {"id": 332, "name": "Drawer", "freebase_id": "/m/0fqfqc"},
+    {"id": 333, "name": "Treadmill", "freebase_id": "/m/030610"},
+    {"id": 334, "name": "Picnic basket", "freebase_id": "/m/07kng9"},
+    {"id": 335, "name": "Dice", "freebase_id": "/m/029b3"},
+    {"id": 336, "name": "Cabbage", "freebase_id": "/m/0fbw6"},
+    {"id": 337, "name": "Football helmet", "freebase_id": "/m/07qxg_"},
+    {"id": 338, "name": "Pig", "freebase_id": "/m/068zj"},
+    {"id": 339, "name": "Person", "freebase_id": "/m/01g317"},
+    {"id": 340, "name": "Shorts", "freebase_id": "/m/01bfm9"},
+    {"id": 341, "name": "Gondola", "freebase_id": "/m/02068x"},
+    {"id": 342, "name": "Honeycomb", "freebase_id": "/m/0fz0h"},
+    {"id": 343, "name": "Doughnut", "freebase_id": "/m/0jy4k"},
+    {"id": 344, "name": "Chest of drawers", "freebase_id": "/m/05kyg_"},
+    {"id": 345, "name": "Land vehicle", "freebase_id": "/m/01prls"},
+    {"id": 346, "name": "Bat", "freebase_id": "/m/01h44"},
+    {"id": 347, "name": "Monkey", "freebase_id": "/m/08pbxl"},
+    {"id": 348, "name": "Dagger", "freebase_id": "/m/02gzp"},
+    {"id": 349, "name": "Tableware", "freebase_id": "/m/04brg2"},
+    {"id": 350, "name": "Human foot", "freebase_id": "/m/031n1"},
+    {"id": 351, "name": "Mug", "freebase_id": "/m/02jvh9"},
+    {"id": 352, "name": "Alarm clock", "freebase_id": "/m/046dlr"},
+    {"id": 353, "name": "Pressure cooker", "freebase_id": "/m/0h8ntjv"},
+    {"id": 354, "name": "Human hand", "freebase_id": "/m/0k65p"},
+    {"id": 355, "name": "Tortoise", "freebase_id": "/m/011k07"},
+    {"id": 356, "name": "Baseball glove", "freebase_id": "/m/03grzl"},
+    {"id": 357, "name": "Sword", "freebase_id": "/m/06y5r"},
+    {"id": 358, "name": "Pear", "freebase_id": "/m/061_f"},
+    {"id": 359, "name": "Miniskirt", "freebase_id": "/m/01cmb2"},
+    {"id": 360, "name": "Traffic sign", "freebase_id": "/m/01mqdt"},
+    {"id": 361, "name": "Girl", "freebase_id": "/m/05r655"},
+    {"id": 362, "name": "Roller skates", "freebase_id": "/m/02p3w7d"},
+    {"id": 363, "name": "Dinosaur", "freebase_id": "/m/029tx"},
+    {"id": 364, "name": "Porch", "freebase_id": "/m/04m6gz"},
+    {"id": 365, "name": "Human beard", "freebase_id": "/m/015h_t"},
+    {"id": 366, "name": "Submarine sandwich", "freebase_id": "/m/06pcq"},
+    {"id": 367, "name": "Screwdriver", "freebase_id": "/m/01bms0"},
+    {"id": 368, "name": "Strawberry", "freebase_id": "/m/07fbm7"},
+    {"id": 369, "name": "Wine glass", "freebase_id": "/m/09tvcd"},
+    {"id": 370, "name": "Seafood", "freebase_id": "/m/06nwz"},
+    {"id": 371, "name": "Racket", "freebase_id": "/m/0dv9c"},
+    {"id": 372, "name": "Wheel", "freebase_id": "/m/083wq"},
+    {"id": 373, "name": "Sea lion", "freebase_id": "/m/0gd36"},
+    {"id": 374, "name": "Toy", "freebase_id": "/m/0138tl"},
+    {"id": 375, "name": "Tea", "freebase_id": "/m/07clx"},
+    {"id": 376, "name": "Tennis ball", "freebase_id": "/m/05ctyq"},
+    {"id": 377, "name": "Waste container", "freebase_id": "/m/0bjyj5"},
+    {"id": 378, "name": "Mule", "freebase_id": "/m/0dbzx"},
+    {"id": 379, "name": "Cricket ball", "freebase_id": "/m/02ctlc"},
+    {"id": 380, "name": "Pineapple", "freebase_id": "/m/0fp6w"},
+    {"id": 381, "name": "Coconut", "freebase_id": "/m/0djtd"},
+    {"id": 382, "name": "Doll", "freebase_id": "/m/0167gd"},
+    {"id": 383, "name": "Coffee table", "freebase_id": "/m/078n6m"},
+    {"id": 384, "name": "Snowman", "freebase_id": "/m/0152hh"},
+    {"id": 385, "name": "Lavender", "freebase_id": "/m/04gth"},
+    {"id": 386, "name": "Shrimp", "freebase_id": "/m/0ll1f78"},
+    {"id": 387, "name": "Maple", "freebase_id": "/m/0cffdh"},
+    {"id": 388, "name": "Cowboy hat", "freebase_id": "/m/025rp__"},
+    {"id": 389, "name": "Goggles", "freebase_id": "/m/02_n6y"},
+    {"id": 390, "name": "Rugby ball", "freebase_id": "/m/0wdt60w"},
+    {"id": 391, "name": "Caterpillar", "freebase_id": "/m/0cydv"},
+    {"id": 392, "name": "Poster", "freebase_id": "/m/01n5jq"},
+    {"id": 393, "name": "Rocket", "freebase_id": "/m/09rvcxw"},
+    {"id": 394, "name": "Organ", "freebase_id": "/m/013y1f"},
+    {"id": 395, "name": "Saxophone", "freebase_id": "/m/06ncr"},
+    {"id": 396, "name": "Traffic light", "freebase_id": "/m/015qff"},
+    {"id": 397, "name": "Cocktail", "freebase_id": "/m/024g6"},
+    {"id": 398, "name": "Plastic bag", "freebase_id": "/m/05gqfk"},
+    {"id": 399, "name": "Squash", "freebase_id": "/m/0dv77"},
+    {"id": 400, "name": "Mushroom", "freebase_id": "/m/052sf"},
+    {"id": 401, "name": "Hamburger", "freebase_id": "/m/0cdn1"},
+    {"id": 402, "name": "Light switch", "freebase_id": "/m/03jbxj"},
+    {"id": 403, "name": "Parachute", "freebase_id": "/m/0cyfs"},
+    {"id": 404, "name": "Teddy bear", "freebase_id": "/m/0kmg4"},
+    {"id": 405, "name": "Winter melon", "freebase_id": "/m/02cvgx"},
+    {"id": 406, "name": "Deer", "freebase_id": "/m/09kx5"},
+    {"id": 407, "name": "Musical keyboard", "freebase_id": "/m/057cc"},
+    {"id": 408, "name": "Plumbing fixture", "freebase_id": "/m/02pkr5"},
+    {"id": 409, "name": "Scoreboard", "freebase_id": "/m/057p5t"},
+    {"id": 410, "name": "Baseball bat", "freebase_id": "/m/03g8mr"},
+    {"id": 411, "name": "Envelope", "freebase_id": "/m/0frqm"},
+    {"id": 412, "name": "Adhesive tape", "freebase_id": "/m/03m3vtv"},
+    {"id": 413, "name": "Briefcase", "freebase_id": "/m/0584n8"},
+    {"id": 414, "name": "Paddle", "freebase_id": "/m/014y4n"},
+    {"id": 415, "name": "Bow and arrow", "freebase_id": "/m/01g3x7"},
+    {"id": 416, "name": "Telephone", "freebase_id": "/m/07cx4"},
+    {"id": 417, "name": "Sheep", "freebase_id": "/m/07bgp"},
+    {"id": 418, "name": "Jacket", "freebase_id": "/m/032b3c"},
+    {"id": 419, "name": "Boy", "freebase_id": "/m/01bl7v"},
+    {"id": 420, "name": "Pizza", "freebase_id": "/m/0663v"},
+    {"id": 421, "name": "Otter", "freebase_id": "/m/0cn6p"},
+    {"id": 422, "name": "Office supplies", "freebase_id": "/m/02rdsp"},
+    {"id": 423, "name": "Couch", "freebase_id": "/m/02crq1"},
+    {"id": 424, "name": "Cello", "freebase_id": "/m/01xqw"},
+    {"id": 425, "name": "Bull", "freebase_id": "/m/0cnyhnx"},
+    {"id": 426, "name": "Camel", "freebase_id": "/m/01x_v"},
+    {"id": 427, "name": "Ball", "freebase_id": "/m/018xm"},
+    {"id": 428, "name": "Duck", "freebase_id": "/m/09ddx"},
+    {"id": 429, "name": "Whale", "freebase_id": "/m/084zz"},
+    {"id": 430, "name": "Shirt", "freebase_id": "/m/01n4qj"},
+    {"id": 431, "name": "Tank", "freebase_id": "/m/07cmd"},
+    {"id": 432, "name": "Motorcycle", "freebase_id": "/m/04_sv"},
+    {"id": 433, "name": "Accordion", "freebase_id": "/m/0mkg"},
+    {"id": 434, "name": "Owl", "freebase_id": "/m/09d5_"},
+    {"id": 435, "name": "Porcupine", "freebase_id": "/m/0c568"},
+    {"id": 436, "name": "Sun hat", "freebase_id": "/m/02wbtzl"},
+    {"id": 437, "name": "Nail", "freebase_id": "/m/05bm6"},
+    {"id": 438, "name": "Scissors", "freebase_id": "/m/01lsmm"},
+    {"id": 439, "name": "Swan", "freebase_id": "/m/0dftk"},
+    {"id": 440, "name": "Lamp", "freebase_id": "/m/0dtln"},
+    {"id": 441, "name": "Crown", "freebase_id": "/m/0nl46"},
+    {"id": 442, "name": "Piano", "freebase_id": "/m/05r5c"},
+    {"id": 443, "name": "Sculpture", "freebase_id": "/m/06msq"},
+    {"id": 444, "name": "Cheetah", "freebase_id": "/m/0cd4d"},
+    {"id": 445, "name": "Oboe", "freebase_id": "/m/05kms"},
+    {"id": 446, "name": "Tin can", "freebase_id": "/m/02jnhm"},
+    {"id": 447, "name": "Mango", "freebase_id": "/m/0fldg"},
+    {"id": 448, "name": "Tripod", "freebase_id": "/m/073bxn"},
+    {"id": 449, "name": "Oven", "freebase_id": "/m/029bxz"},
+    {"id": 450, "name": "Mouse", "freebase_id": "/m/020lf"},
+    {"id": 451, "name": "Barge", "freebase_id": "/m/01btn"},
+    {"id": 452, "name": "Coffee", "freebase_id": "/m/02vqfm"},
+    {"id": 453, "name": "Snowboard", "freebase_id": "/m/06__v"},
+    {"id": 454, "name": "Common fig", "freebase_id": "/m/043nyj"},
+    {"id": 455, "name": "Salad", "freebase_id": "/m/0grw1"},
+    {"id": 456, "name": "Marine invertebrates", "freebase_id": "/m/03hl4l9"},
+    {"id": 457, "name": "Umbrella", "freebase_id": "/m/0hnnb"},
+    {"id": 458, "name": "Kangaroo", "freebase_id": "/m/04c0y"},
+    {"id": 459, "name": "Human arm", "freebase_id": "/m/0dzf4"},
+    {"id": 460, "name": "Measuring cup", "freebase_id": "/m/07v9_z"},
+    {"id": 461, "name": "Snail", "freebase_id": "/m/0f9_l"},
+    {"id": 462, "name": "Loveseat", "freebase_id": "/m/0703r8"},
+    {"id": 463, "name": "Suit", "freebase_id": "/m/01xyhv"},
+    {"id": 464, "name": "Teapot", "freebase_id": "/m/01fh4r"},
+    {"id": 465, "name": "Bottle", "freebase_id": "/m/04dr76w"},
+    {"id": 466, "name": "Alpaca", "freebase_id": "/m/0pcr"},
+    {"id": 467, "name": "Kettle", "freebase_id": "/m/03s_tn"},
+    {"id": 468, "name": "Trousers", "freebase_id": "/m/07mhn"},
+    {"id": 469, "name": "Popcorn", "freebase_id": "/m/01hrv5"},
+    {"id": 470, "name": "Centipede", "freebase_id": "/m/019h78"},
+    {"id": 471, "name": "Spider", "freebase_id": "/m/09kmb"},
+    {"id": 472, "name": "Sparrow", "freebase_id": "/m/0h23m"},
+    {"id": 473, "name": "Plate", "freebase_id": "/m/050gv4"},
+    {"id": 474, "name": "Bagel", "freebase_id": "/m/01fb_0"},
+    {"id": 475, "name": "Personal care", "freebase_id": "/m/02w3_ws"},
+    {"id": 476, "name": "Apple", "freebase_id": "/m/014j1m"},
+    {"id": 477, "name": "Brassiere", "freebase_id": "/m/01gmv2"},
+    {"id": 478, "name": "Bathroom cabinet", "freebase_id": "/m/04y4h8h"},
+    {"id": 479, "name": "studio couch", "freebase_id": "/m/026qbn5"},
+    {"id": 480, "name": "Computer keyboard", "freebase_id": "/m/01m2v"},
+    {"id": 481, "name": "Table tennis racket", "freebase_id": "/m/05_5p_0"},
+    {"id": 482, "name": "Sushi", "freebase_id": "/m/07030"},
+    {"id": 483, "name": "Cabinetry", "freebase_id": "/m/01s105"},
+    {"id": 484, "name": "Street light", "freebase_id": "/m/033rq4"},
+    {"id": 485, "name": "Towel", "freebase_id": "/m/0162_1"},
+    {"id": 486, "name": "Nightstand", "freebase_id": "/m/02z51p"},
+    {"id": 487, "name": "Rabbit", "freebase_id": "/m/06mf6"},
+    {"id": 488, "name": "Dolphin", "freebase_id": "/m/02hj4"},
+    {"id": 489, "name": "Dog", "freebase_id": "/m/0bt9lr"},
+    {"id": 490, "name": "Jug", "freebase_id": "/m/08hvt4"},
+    {"id": 491, "name": "Wok", "freebase_id": "/m/084rd"},
+    {"id": 492, "name": "Fire hydrant", "freebase_id": "/m/01pns0"},
+    {"id": 493, "name": "Human eye", "freebase_id": "/m/014sv8"},
+    {"id": 494, "name": "Skyscraper", "freebase_id": "/m/079cl"},
+    {"id": 495, "name": "Backpack", "freebase_id": "/m/01940j"},
+    {"id": 496, "name": "Potato", "freebase_id": "/m/05vtc"},
+    {"id": 497, "name": "Paper towel", "freebase_id": "/m/02w3r3"},
+    {"id": 498, "name": "Lifejacket", "freebase_id": "/m/054xkw"},
+    {"id": 499, "name": "Bicycle wheel", "freebase_id": "/m/01bqk0"},
+    {"id": 500, "name": "Toilet", "freebase_id": "/m/09g1w"},
+]
+
+
+OPENIMAGES_V6_CATEGORIES = [
+    {"id": 1, "name": "Tortoise", "freebase_id": "/m/011k07"},
+    {"id": 2, "name": "Container", "freebase_id": "/m/011q46kg"},
+    {"id": 3, "name": "Magpie", "freebase_id": "/m/012074"},
+    {"id": 4, "name": "Sea turtle", "freebase_id": "/m/0120dh"},
+    {"id": 5, "name": "Football", "freebase_id": "/m/01226z"},
+    {"id": 6, "name": "Ambulance", "freebase_id": "/m/012n7d"},
+    {"id": 7, "name": "Ladder", "freebase_id": "/m/012w5l"},
+    {"id": 8, "name": "Toothbrush", "freebase_id": "/m/012xff"},
+    {"id": 9, "name": "Syringe", "freebase_id": "/m/012ysf"},
+    {"id": 10, "name": "Sink", "freebase_id": "/m/0130jx"},
+    {"id": 11, "name": "Toy", "freebase_id": "/m/0138tl"},
+    {"id": 12, "name": "Organ (Musical Instrument)", "freebase_id": "/m/013y1f"},
+    {"id": 13, "name": "Cassette deck", "freebase_id": "/m/01432t"},
+    {"id": 14, "name": "Apple", "freebase_id": "/m/014j1m"},
+    {"id": 15, "name": "Human eye", "freebase_id": "/m/014sv8"},
+    {"id": 16, "name": "Cosmetics", "freebase_id": "/m/014trl"},
+    {"id": 17, "name": "Paddle", "freebase_id": "/m/014y4n"},
+    {"id": 18, "name": "Snowman", "freebase_id": "/m/0152hh"},
+    {"id": 19, "name": "Beer", "freebase_id": "/m/01599"},
+    {"id": 20, "name": "Chopsticks", "freebase_id": "/m/01_5g"},
+    {"id": 21, "name": "Human beard", "freebase_id": "/m/015h_t"},
+    {"id": 22, "name": "Bird", "freebase_id": "/m/015p6"},
+    {"id": 23, "name": "Parking meter", "freebase_id": "/m/015qbp"},
+    {"id": 24, "name": "Traffic light", "freebase_id": "/m/015qff"},
+    {"id": 25, "name": "Croissant", "freebase_id": "/m/015wgc"},
+    {"id": 26, "name": "Cucumber", "freebase_id": "/m/015x4r"},
+    {"id": 27, "name": "Radish", "freebase_id": "/m/015x5n"},
+    {"id": 28, "name": "Towel", "freebase_id": "/m/0162_1"},
+    {"id": 29, "name": "Doll", "freebase_id": "/m/0167gd"},
+    {"id": 30, "name": "Skull", "freebase_id": "/m/016m2d"},
+    {"id": 31, "name": "Washing machine", "freebase_id": "/m/0174k2"},
+    {"id": 32, "name": "Glove", "freebase_id": "/m/0174n1"},
+    {"id": 33, "name": "Tick", "freebase_id": "/m/0175cv"},
+    {"id": 34, "name": "Belt", "freebase_id": "/m/0176mf"},
+    {"id": 35, "name": "Sunglasses", "freebase_id": "/m/017ftj"},
+    {"id": 36, "name": "Banjo", "freebase_id": "/m/018j2"},
+    {"id": 37, "name": "Cart", "freebase_id": "/m/018p4k"},
+    {"id": 38, "name": "Ball", "freebase_id": "/m/018xm"},
+    {"id": 39, "name": "Backpack", "freebase_id": "/m/01940j"},
+    {"id": 40, "name": "Bicycle", "freebase_id": "/m/0199g"},
+    {"id": 41, "name": "Home appliance", "freebase_id": "/m/019dx1"},
+    {"id": 42, "name": "Centipede", "freebase_id": "/m/019h78"},
+    {"id": 43, "name": "Boat", "freebase_id": "/m/019jd"},
+    {"id": 44, "name": "Surfboard", "freebase_id": "/m/019w40"},
+    {"id": 45, "name": "Boot", "freebase_id": "/m/01b638"},
+    {"id": 46, "name": "Headphones", "freebase_id": "/m/01b7fy"},
+    {"id": 47, "name": "Hot dog", "freebase_id": "/m/01b9xk"},
+    {"id": 48, "name": "Shorts", "freebase_id": "/m/01bfm9"},
+    {"id": 49, "name": "Fast food", "freebase_id": "/m/01_bhs"},
+    {"id": 50, "name": "Bus", "freebase_id": "/m/01bjv"},
+    {"id": 51, "name": "Boy", "freebase_id": "/m/01bl7v"},
+    {"id": 52, "name": "Screwdriver", "freebase_id": "/m/01bms0"},
+    {"id": 53, "name": "Bicycle wheel", "freebase_id": "/m/01bqk0"},
+    {"id": 54, "name": "Barge", "freebase_id": "/m/01btn"},
+    {"id": 55, "name": "Laptop", "freebase_id": "/m/01c648"},
+    {"id": 56, "name": "Miniskirt", "freebase_id": "/m/01cmb2"},
+    {"id": 57, "name": "Drill (Tool)", "freebase_id": "/m/01d380"},
+    {"id": 58, "name": "Dress", "freebase_id": "/m/01d40f"},
+    {"id": 59, "name": "Bear", "freebase_id": "/m/01dws"},
+    {"id": 60, "name": "Waffle", "freebase_id": "/m/01dwsz"},
+    {"id": 61, "name": "Pancake", "freebase_id": "/m/01dwwc"},
+    {"id": 62, "name": "Brown bear", "freebase_id": "/m/01dxs"},
+    {"id": 63, "name": "Woodpecker", "freebase_id": "/m/01dy8n"},
+    {"id": 64, "name": "Blue jay", "freebase_id": "/m/01f8m5"},
+    {"id": 65, "name": "Pretzel", "freebase_id": "/m/01f91_"},
+    {"id": 66, "name": "Bagel", "freebase_id": "/m/01fb_0"},
+    {"id": 67, "name": "Tower", "freebase_id": "/m/01fdzj"},
+    {"id": 68, "name": "Teapot", "freebase_id": "/m/01fh4r"},
+    {"id": 69, "name": "Person", "freebase_id": "/m/01g317"},
+    {"id": 70, "name": "Bow and arrow", "freebase_id": "/m/01g3x7"},
+    {"id": 71, "name": "Swimwear", "freebase_id": "/m/01gkx_"},
+    {"id": 72, "name": "Beehive", "freebase_id": "/m/01gllr"},
+    {"id": 73, "name": "Brassiere", "freebase_id": "/m/01gmv2"},
+    {"id": 74, "name": "Bee", "freebase_id": "/m/01h3n"},
+    {"id": 75, "name": "Bat (Animal)", "freebase_id": "/m/01h44"},
+    {"id": 76, "name": "Starfish", "freebase_id": "/m/01h8tj"},
+    {"id": 77, "name": "Popcorn", "freebase_id": "/m/01hrv5"},
+    {"id": 78, "name": "Burrito", "freebase_id": "/m/01j3zr"},
+    {"id": 79, "name": "Chainsaw", "freebase_id": "/m/01j4z9"},
+    {"id": 80, "name": "Balloon", "freebase_id": "/m/01j51"},
+    {"id": 81, "name": "Wrench", "freebase_id": "/m/01j5ks"},
+    {"id": 82, "name": "Tent", "freebase_id": "/m/01j61q"},
+    {"id": 83, "name": "Vehicle registration plate", "freebase_id": "/m/01jfm_"},
+    {"id": 84, "name": "Lantern", "freebase_id": "/m/01jfsr"},
+    {"id": 85, "name": "Toaster", "freebase_id": "/m/01k6s3"},
+    {"id": 86, "name": "Flashlight", "freebase_id": "/m/01kb5b"},
+    {"id": 87, "name": "Billboard", "freebase_id": "/m/01knjb"},
+    {"id": 88, "name": "Tiara", "freebase_id": "/m/01krhy"},
+    {"id": 89, "name": "Limousine", "freebase_id": "/m/01lcw4"},
+    {"id": 90, "name": "Necklace", "freebase_id": "/m/01llwg"},
+    {"id": 91, "name": "Carnivore", "freebase_id": "/m/01lrl"},
+    {"id": 92, "name": "Scissors", "freebase_id": "/m/01lsmm"},
+    {"id": 93, "name": "Stairs", "freebase_id": "/m/01lynh"},
+    {"id": 94, "name": "Computer keyboard", "freebase_id": "/m/01m2v"},
+    {"id": 95, "name": "Printer", "freebase_id": "/m/01m4t"},
+    {"id": 96, "name": "Traffic sign", "freebase_id": "/m/01mqdt"},
+    {"id": 97, "name": "Chair", "freebase_id": "/m/01mzpv"},
+    {"id": 98, "name": "Shirt", "freebase_id": "/m/01n4qj"},
+    {"id": 99, "name": "Poster", "freebase_id": "/m/01n5jq"},
+    {"id": 100, "name": "Cheese", "freebase_id": "/m/01nkt"},
+    {"id": 101, "name": "Sock", "freebase_id": "/m/01nq26"},
+    {"id": 102, "name": "Fire hydrant", "freebase_id": "/m/01pns0"},
+    {"id": 103, "name": "Land vehicle", "freebase_id": "/m/01prls"},
+    {"id": 104, "name": "Earrings", "freebase_id": "/m/01r546"},
+    {"id": 105, "name": "Tie", "freebase_id": "/m/01rkbr"},
+    {"id": 106, "name": "Watercraft", "freebase_id": "/m/01rzcn"},
+    {"id": 107, "name": "Cabinetry", "freebase_id": "/m/01s105"},
+    {"id": 108, "name": "Suitcase", "freebase_id": "/m/01s55n"},
+    {"id": 109, "name": "Muffin", "freebase_id": "/m/01tcjp"},
+    {"id": 110, "name": "Bidet", "freebase_id": "/m/01vbnl"},
+    {"id": 111, "name": "Snack", "freebase_id": "/m/01ww8y"},
+    {"id": 112, "name": "Snowmobile", "freebase_id": "/m/01x3jk"},
+    {"id": 113, "name": "Clock", "freebase_id": "/m/01x3z"},
+    {"id": 114, "name": "Medical equipment", "freebase_id": "/m/01xgg_"},
+    {"id": 115, "name": "Cattle", "freebase_id": "/m/01xq0k1"},
+    {"id": 116, "name": "Cello", "freebase_id": "/m/01xqw"},
+    {"id": 117, "name": "Jet ski", "freebase_id": "/m/01xs3r"},
+    {"id": 118, "name": "Camel", "freebase_id": "/m/01x_v"},
+    {"id": 119, "name": "Coat", "freebase_id": "/m/01xygc"},
+    {"id": 120, "name": "Suit", "freebase_id": "/m/01xyhv"},
+    {"id": 121, "name": "Desk", "freebase_id": "/m/01y9k5"},
+    {"id": 122, "name": "Cat", "freebase_id": "/m/01yrx"},
+    {"id": 123, "name": "Bronze sculpture", "freebase_id": "/m/01yx86"},
+    {"id": 124, "name": "Juice", "freebase_id": "/m/01z1kdw"},
+    {"id": 125, "name": "Gondola", "freebase_id": "/m/02068x"},
+    {"id": 126, "name": "Beetle", "freebase_id": "/m/020jm"},
+    {"id": 127, "name": "Cannon", "freebase_id": "/m/020kz"},
+    {"id": 128, "name": "Computer mouse", "freebase_id": "/m/020lf"},
+    {"id": 129, "name": "Cookie", "freebase_id": "/m/021mn"},
+    {"id": 130, "name": "Office building", "freebase_id": "/m/021sj1"},
+    {"id": 131, "name": "Fountain", "freebase_id": "/m/0220r2"},
+    {"id": 132, "name": "Coin", "freebase_id": "/m/0242l"},
+    {"id": 133, "name": "Calculator", "freebase_id": "/m/024d2"},
+    {"id": 134, "name": "Cocktail", "freebase_id": "/m/024g6"},
+    {"id": 135, "name": "Computer monitor", "freebase_id": "/m/02522"},
+    {"id": 136, "name": "Box", "freebase_id": "/m/025dyy"},
+    {"id": 137, "name": "Stapler", "freebase_id": "/m/025fsf"},
+    {"id": 138, "name": "Christmas tree", "freebase_id": "/m/025nd"},
+    {"id": 139, "name": "Cowboy hat", "freebase_id": "/m/025rp__"},
+    {"id": 140, "name": "Hiking equipment", "freebase_id": "/m/0268lbt"},
+    {"id": 141, "name": "Studio couch", "freebase_id": "/m/026qbn5"},
+    {"id": 142, "name": "Drum", "freebase_id": "/m/026t6"},
+    {"id": 143, "name": "Dessert", "freebase_id": "/m/0270h"},
+    {"id": 144, "name": "Wine rack", "freebase_id": "/m/0271qf7"},
+    {"id": 145, "name": "Drink", "freebase_id": "/m/0271t"},
+    {"id": 146, "name": "Zucchini", "freebase_id": "/m/027pcv"},
+    {"id": 147, "name": "Ladle", "freebase_id": "/m/027rl48"},
+    {"id": 148, "name": "Human mouth", "freebase_id": "/m/0283dt1"},
+    {"id": 149, "name": "Dairy Product", "freebase_id": "/m/0284d"},
+    {"id": 150, "name": "Dice", "freebase_id": "/m/029b3"},
+    {"id": 151, "name": "Oven", "freebase_id": "/m/029bxz"},
+    {"id": 152, "name": "Dinosaur", "freebase_id": "/m/029tx"},
+    {"id": 153, "name": "Ratchet (Device)", "freebase_id": "/m/02bm9n"},
+    {"id": 154, "name": "Couch", "freebase_id": "/m/02crq1"},
+    {"id": 155, "name": "Cricket ball", "freebase_id": "/m/02ctlc"},
+    {"id": 156, "name": "Winter melon", "freebase_id": "/m/02cvgx"},
+    {"id": 157, "name": "Spatula", "freebase_id": "/m/02d1br"},
+    {"id": 158, "name": "Whiteboard", "freebase_id": "/m/02d9qx"},
+    {"id": 159, "name": "Pencil sharpener", "freebase_id": "/m/02ddwp"},
+    {"id": 160, "name": "Door", "freebase_id": "/m/02dgv"},
+    {"id": 161, "name": "Hat", "freebase_id": "/m/02dl1y"},
+    {"id": 162, "name": "Shower", "freebase_id": "/m/02f9f_"},
+    {"id": 163, "name": "Eraser", "freebase_id": "/m/02fh7f"},
+    {"id": 164, "name": "Fedora", "freebase_id": "/m/02fq_6"},
+    {"id": 165, "name": "Guacamole", "freebase_id": "/m/02g30s"},
+    {"id": 166, "name": "Dagger", "freebase_id": "/m/02gzp"},
+    {"id": 167, "name": "Scarf", "freebase_id": "/m/02h19r"},
+    {"id": 168, "name": "Dolphin", "freebase_id": "/m/02hj4"},
+    {"id": 169, "name": "Sombrero", "freebase_id": "/m/02jfl0"},
+    {"id": 170, "name": "Tin can", "freebase_id": "/m/02jnhm"},
+    {"id": 171, "name": "Mug", "freebase_id": "/m/02jvh9"},
+    {"id": 172, "name": "Tap", "freebase_id": "/m/02jz0l"},
+    {"id": 173, "name": "Harbor seal", "freebase_id": "/m/02l8p9"},
+    {"id": 174, "name": "Stretcher", "freebase_id": "/m/02lbcq"},
+    {"id": 175, "name": "Can opener", "freebase_id": "/m/02mqfb"},
+    {"id": 176, "name": "Goggles", "freebase_id": "/m/02_n6y"},
+    {"id": 177, "name": "Human body", "freebase_id": "/m/02p0tk3"},
+    {"id": 178, "name": "Roller skates", "freebase_id": "/m/02p3w7d"},
+    {"id": 179, "name": "Coffee cup", "freebase_id": "/m/02p5f1q"},
+    {"id": 180, "name": "Cutting board", "freebase_id": "/m/02pdsw"},
+    {"id": 181, "name": "Blender", "freebase_id": "/m/02pjr4"},
+    {"id": 182, "name": "Plumbing fixture", "freebase_id": "/m/02pkr5"},
+    {"id": 183, "name": "Stop sign", "freebase_id": "/m/02pv19"},
+    {"id": 184, "name": "Office supplies", "freebase_id": "/m/02rdsp"},
+    {"id": 185, "name": "Volleyball (Ball)", "freebase_id": "/m/02rgn06"},
+    {"id": 186, "name": "Vase", "freebase_id": "/m/02s195"},
+    {"id": 187, "name": "Slow cooker", "freebase_id": "/m/02tsc9"},
+    {"id": 188, "name": "Wardrobe", "freebase_id": "/m/02vkqh8"},
+    {"id": 189, "name": "Coffee", "freebase_id": "/m/02vqfm"},
+    {"id": 190, "name": "Whisk", "freebase_id": "/m/02vwcm"},
+    {"id": 191, "name": "Paper towel", "freebase_id": "/m/02w3r3"},
+    {"id": 192, "name": "Personal care", "freebase_id": "/m/02w3_ws"},
+    {"id": 193, "name": "Food", "freebase_id": "/m/02wbm"},
+    {"id": 194, "name": "Sun hat", "freebase_id": "/m/02wbtzl"},
+    {"id": 195, "name": "Tree house", "freebase_id": "/m/02wg_p"},
+    {"id": 196, "name": "Flying disc", "freebase_id": "/m/02wmf"},
+    {"id": 197, "name": "Skirt", "freebase_id": "/m/02wv6h6"},
+    {"id": 198, "name": "Gas stove", "freebase_id": "/m/02wv84t"},
+    {"id": 199, "name": "Salt and pepper shakers", "freebase_id": "/m/02x8cch"},
+    {"id": 200, "name": "Mechanical fan", "freebase_id": "/m/02x984l"},
+    {"id": 201, "name": "Face powder", "freebase_id": "/m/02xb7qb"},
+    {"id": 202, "name": "Fax", "freebase_id": "/m/02xqq"},
+    {"id": 203, "name": "Fruit", "freebase_id": "/m/02xwb"},
+    {"id": 204, "name": "French fries", "freebase_id": "/m/02y6n"},
+    {"id": 205, "name": "Nightstand", "freebase_id": "/m/02z51p"},
+    {"id": 206, "name": "Barrel", "freebase_id": "/m/02zn6n"},
+    {"id": 207, "name": "Kite", "freebase_id": "/m/02zt3"},
+    {"id": 208, "name": "Tart", "freebase_id": "/m/02zvsm"},
+    {"id": 209, "name": "Treadmill", "freebase_id": "/m/030610"},
+    {"id": 210, "name": "Fox", "freebase_id": "/m/0306r"},
+    {"id": 211, "name": "Flag", "freebase_id": "/m/03120"},
+    {"id": 212, "name": "French horn", "freebase_id": "/m/0319l"},
+    {"id": 213, "name": "Window blind", "freebase_id": "/m/031b6r"},
+    {"id": 214, "name": "Human foot", "freebase_id": "/m/031n1"},
+    {"id": 215, "name": "Golf cart", "freebase_id": "/m/0323sq"},
+    {"id": 216, "name": "Jacket", "freebase_id": "/m/032b3c"},
+    {"id": 217, "name": "Egg (Food)", "freebase_id": "/m/033cnk"},
+    {"id": 218, "name": "Street light", "freebase_id": "/m/033rq4"},
+    {"id": 219, "name": "Guitar", "freebase_id": "/m/0342h"},
+    {"id": 220, "name": "Pillow", "freebase_id": "/m/034c16"},
+    {"id": 221, "name": "Human leg", "freebase_id": "/m/035r7c"},
+    {"id": 222, "name": "Isopod", "freebase_id": "/m/035vxb"},
+    {"id": 223, "name": "Grape", "freebase_id": "/m/0388q"},
+    {"id": 224, "name": "Human ear", "freebase_id": "/m/039xj_"},
+    {"id": 225, "name": "Power plugs and sockets", "freebase_id": "/m/03bbps"},
+    {"id": 226, "name": "Panda", "freebase_id": "/m/03bj1"},
+    {"id": 227, "name": "Giraffe", "freebase_id": "/m/03bk1"},
+    {"id": 228, "name": "Woman", "freebase_id": "/m/03bt1vf"},
+    {"id": 229, "name": "Door handle", "freebase_id": "/m/03c7gz"},
+    {"id": 230, "name": "Rhinoceros", "freebase_id": "/m/03d443"},
+    {"id": 231, "name": "Bathtub", "freebase_id": "/m/03dnzn"},
+    {"id": 232, "name": "Goldfish", "freebase_id": "/m/03fj2"},
+    {"id": 233, "name": "Houseplant", "freebase_id": "/m/03fp41"},
+    {"id": 234, "name": "Goat", "freebase_id": "/m/03fwl"},
+    {"id": 235, "name": "Baseball bat", "freebase_id": "/m/03g8mr"},
+    {"id": 236, "name": "Baseball glove", "freebase_id": "/m/03grzl"},
+    {"id": 237, "name": "Mixing bowl", "freebase_id": "/m/03hj559"},
+    {"id": 238, "name": "Marine invertebrates", "freebase_id": "/m/03hl4l9"},
+    {"id": 239, "name": "Kitchen utensil", "freebase_id": "/m/03hlz0c"},
+    {"id": 240, "name": "Light switch", "freebase_id": "/m/03jbxj"},
+    {"id": 241, "name": "House", "freebase_id": "/m/03jm5"},
+    {"id": 242, "name": "Horse", "freebase_id": "/m/03k3r"},
+    {"id": 243, "name": "Stationary bicycle", "freebase_id": "/m/03kt2w"},
+    {"id": 244, "name": "Hammer", "freebase_id": "/m/03l9g"},
+    {"id": 245, "name": "Ceiling fan", "freebase_id": "/m/03ldnb"},
+    {"id": 246, "name": "Sofa bed", "freebase_id": "/m/03m3pdh"},
+    {"id": 247, "name": "Adhesive tape", "freebase_id": "/m/03m3vtv"},
+    {"id": 248, "name": "Harp", "freebase_id": "/m/03m5k"},
+    {"id": 249, "name": "Sandal", "freebase_id": "/m/03nfch"},
+    {"id": 250, "name": "Bicycle helmet", "freebase_id": "/m/03p3bw"},
+    {"id": 251, "name": "Saucer", "freebase_id": "/m/03q5c7"},
+    {"id": 252, "name": "Harpsichord", "freebase_id": "/m/03q5t"},
+    {"id": 253, "name": "Human hair", "freebase_id": "/m/03q69"},
+    {"id": 254, "name": "Heater", "freebase_id": "/m/03qhv5"},
+    {"id": 255, "name": "Harmonica", "freebase_id": "/m/03qjg"},
+    {"id": 256, "name": "Hamster", "freebase_id": "/m/03qrc"},
+    {"id": 257, "name": "Curtain", "freebase_id": "/m/03rszm"},
+    {"id": 258, "name": "Bed", "freebase_id": "/m/03ssj5"},
+    {"id": 259, "name": "Kettle", "freebase_id": "/m/03s_tn"},
+    {"id": 260, "name": "Fireplace", "freebase_id": "/m/03tw93"},
+    {"id": 261, "name": "Scale", "freebase_id": "/m/03txqz"},
+    {"id": 262, "name": "Drinking straw", "freebase_id": "/m/03v5tg"},
+    {"id": 263, "name": "Insect", "freebase_id": "/m/03vt0"},
+    {"id": 264, "name": "Hair dryer", "freebase_id": "/m/03wvsk"},
+    {"id": 265, "name": "Kitchenware", "freebase_id": "/m/03_wxk"},
+    {"id": 266, "name": "Indoor rower", "freebase_id": "/m/03wym"},
+    {"id": 267, "name": "Invertebrate", "freebase_id": "/m/03xxp"},
+    {"id": 268, "name": "Food processor", "freebase_id": "/m/03y6mg"},
+    {"id": 269, "name": "Bookcase", "freebase_id": "/m/03__z0"},
+    {"id": 270, "name": "Refrigerator", "freebase_id": "/m/040b_t"},
+    {"id": 271, "name": "Wood-burning stove", "freebase_id": "/m/04169hn"},
+    {"id": 272, "name": "Punching bag", "freebase_id": "/m/0420v5"},
+    {"id": 273, "name": "Common fig", "freebase_id": "/m/043nyj"},
+    {"id": 274, "name": "Cocktail shaker", "freebase_id": "/m/0440zs"},
+    {"id": 275, "name": "Jaguar (Animal)", "freebase_id": "/m/0449p"},
+    {"id": 276, "name": "Golf ball", "freebase_id": "/m/044r5d"},
+    {"id": 277, "name": "Fashion accessory", "freebase_id": "/m/0463sg"},
+    {"id": 278, "name": "Alarm clock", "freebase_id": "/m/046dlr"},
+    {"id": 279, "name": "Filing cabinet", "freebase_id": "/m/047j0r"},
+    {"id": 280, "name": "Artichoke", "freebase_id": "/m/047v4b"},
+    {"id": 281, "name": "Table", "freebase_id": "/m/04bcr3"},
+    {"id": 282, "name": "Tableware", "freebase_id": "/m/04brg2"},
+    {"id": 283, "name": "Kangaroo", "freebase_id": "/m/04c0y"},
+    {"id": 284, "name": "Koala", "freebase_id": "/m/04cp_"},
+    {"id": 285, "name": "Knife", "freebase_id": "/m/04ctx"},
+    {"id": 286, "name": "Bottle", "freebase_id": "/m/04dr76w"},
+    {"id": 287, "name": "Bottle opener", "freebase_id": "/m/04f5ws"},
+    {"id": 288, "name": "Lynx", "freebase_id": "/m/04g2r"},
+    {"id": 289, "name": "Lavender (Plant)", "freebase_id": "/m/04gth"},
+    {"id": 290, "name": "Lighthouse", "freebase_id": "/m/04h7h"},
+    {"id": 291, "name": "Dumbbell", "freebase_id": "/m/04h8sr"},
+    {"id": 292, "name": "Human head", "freebase_id": "/m/04hgtk"},
+    {"id": 293, "name": "Bowl", "freebase_id": "/m/04kkgm"},
+    {"id": 294, "name": "Humidifier", "freebase_id": "/m/04lvq_"},
+    {"id": 295, "name": "Porch", "freebase_id": "/m/04m6gz"},
+    {"id": 296, "name": "Lizard", "freebase_id": "/m/04m9y"},
+    {"id": 297, "name": "Billiard table", "freebase_id": "/m/04p0qw"},
+    {"id": 298, "name": "Mammal", "freebase_id": "/m/04rky"},
+    {"id": 299, "name": "Mouse", "freebase_id": "/m/04rmv"},
+    {"id": 300, "name": "Motorcycle", "freebase_id": "/m/04_sv"},
+    {"id": 301, "name": "Musical instrument", "freebase_id": "/m/04szw"},
+    {"id": 302, "name": "Swim cap", "freebase_id": "/m/04tn4x"},
+    {"id": 303, "name": "Frying pan", "freebase_id": "/m/04v6l4"},
+    {"id": 304, "name": "Snowplow", "freebase_id": "/m/04vv5k"},
+    {"id": 305, "name": "Bathroom cabinet", "freebase_id": "/m/04y4h8h"},
+    {"id": 306, "name": "Missile", "freebase_id": "/m/04ylt"},
+    {"id": 307, "name": "Bust", "freebase_id": "/m/04yqq2"},
+    {"id": 308, "name": "Man", "freebase_id": "/m/04yx4"},
+    {"id": 309, "name": "Waffle iron", "freebase_id": "/m/04z4wx"},
+    {"id": 310, "name": "Milk", "freebase_id": "/m/04zpv"},
+    {"id": 311, "name": "Ring binder", "freebase_id": "/m/04zwwv"},
+    {"id": 312, "name": "Plate", "freebase_id": "/m/050gv4"},
+    {"id": 313, "name": "Mobile phone", "freebase_id": "/m/050k8"},
+    {"id": 314, "name": "Baked goods", "freebase_id": "/m/052lwg6"},
+    {"id": 315, "name": "Mushroom", "freebase_id": "/m/052sf"},
+    {"id": 316, "name": "Crutch", "freebase_id": "/m/05441v"},
+    {"id": 317, "name": "Pitcher (Container)", "freebase_id": "/m/054fyh"},
+    {"id": 318, "name": "Mirror", "freebase_id": "/m/054_l"},
+    {"id": 319, "name": "Personal flotation device", "freebase_id": "/m/054xkw"},
+    {"id": 320, "name": "Table tennis racket", "freebase_id": "/m/05_5p_0"},
+    {"id": 321, "name": "Pencil case", "freebase_id": "/m/05676x"},
+    {"id": 322, "name": "Musical keyboard", "freebase_id": "/m/057cc"},
+    {"id": 323, "name": "Scoreboard", "freebase_id": "/m/057p5t"},
+    {"id": 324, "name": "Briefcase", "freebase_id": "/m/0584n8"},
+    {"id": 325, "name": "Kitchen knife", "freebase_id": "/m/058qzx"},
+    {"id": 326, "name": "Nail (Construction)", "freebase_id": "/m/05bm6"},
+    {"id": 327, "name": "Tennis ball", "freebase_id": "/m/05ctyq"},
+    {"id": 328, "name": "Plastic bag", "freebase_id": "/m/05gqfk"},
+    {"id": 329, "name": "Oboe", "freebase_id": "/m/05kms"},
+    {"id": 330, "name": "Chest of drawers", "freebase_id": "/m/05kyg_"},
+    {"id": 331, "name": "Ostrich", "freebase_id": "/m/05n4y"},
+    {"id": 332, "name": "Piano", "freebase_id": "/m/05r5c"},
+    {"id": 333, "name": "Girl", "freebase_id": "/m/05r655"},
+    {"id": 334, "name": "Plant", "freebase_id": "/m/05s2s"},
+    {"id": 335, "name": "Potato", "freebase_id": "/m/05vtc"},
+    {"id": 336, "name": "Hair spray", "freebase_id": "/m/05w9t9"},
+    {"id": 337, "name": "Sports equipment", "freebase_id": "/m/05y5lj"},
+    {"id": 338, "name": "Pasta", "freebase_id": "/m/05z55"},
+    {"id": 339, "name": "Penguin", "freebase_id": "/m/05z6w"},
+    {"id": 340, "name": "Pumpkin", "freebase_id": "/m/05zsy"},
+    {"id": 341, "name": "Pear", "freebase_id": "/m/061_f"},
+    {"id": 342, "name": "Infant bed", "freebase_id": "/m/061hd_"},
+    {"id": 343, "name": "Polar bear", "freebase_id": "/m/0633h"},
+    {"id": 344, "name": "Mixer", "freebase_id": "/m/063rgb"},
+    {"id": 345, "name": "Cupboard", "freebase_id": "/m/0642b4"},
+    {"id": 346, "name": "Jacuzzi", "freebase_id": "/m/065h6l"},
+    {"id": 347, "name": "Pizza", "freebase_id": "/m/0663v"},
+    {"id": 348, "name": "Digital clock", "freebase_id": "/m/06_72j"},
+    {"id": 349, "name": "Pig", "freebase_id": "/m/068zj"},
+    {"id": 350, "name": "Reptile", "freebase_id": "/m/06bt6"},
+    {"id": 351, "name": "Rifle", "freebase_id": "/m/06c54"},
+    {"id": 352, "name": "Lipstick", "freebase_id": "/m/06c7f7"},
+    {"id": 353, "name": "Skateboard", "freebase_id": "/m/06_fw"},
+    {"id": 354, "name": "Raven", "freebase_id": "/m/06j2d"},
+    {"id": 355, "name": "High heels", "freebase_id": "/m/06k2mb"},
+    {"id": 356, "name": "Red panda", "freebase_id": "/m/06l9r"},
+    {"id": 357, "name": "Rose", "freebase_id": "/m/06m11"},
+    {"id": 358, "name": "Rabbit", "freebase_id": "/m/06mf6"},
+    {"id": 359, "name": "Sculpture", "freebase_id": "/m/06msq"},
+    {"id": 360, "name": "Saxophone", "freebase_id": "/m/06ncr"},
+    {"id": 361, "name": "Shotgun", "freebase_id": "/m/06nrc"},
+    {"id": 362, "name": "Seafood", "freebase_id": "/m/06nwz"},
+    {"id": 363, "name": "Submarine sandwich", "freebase_id": "/m/06pcq"},
+    {"id": 364, "name": "Snowboard", "freebase_id": "/m/06__v"},
+    {"id": 365, "name": "Sword", "freebase_id": "/m/06y5r"},
+    {"id": 366, "name": "Picture frame", "freebase_id": "/m/06z37_"},
+    {"id": 367, "name": "Sushi", "freebase_id": "/m/07030"},
+    {"id": 368, "name": "Loveseat", "freebase_id": "/m/0703r8"},
+    {"id": 369, "name": "Ski", "freebase_id": "/m/071p9"},
+    {"id": 370, "name": "Squirrel", "freebase_id": "/m/071qp"},
+    {"id": 371, "name": "Tripod", "freebase_id": "/m/073bxn"},
+    {"id": 372, "name": "Stethoscope", "freebase_id": "/m/073g6"},
+    {"id": 373, "name": "Submarine", "freebase_id": "/m/074d1"},
+    {"id": 374, "name": "Scorpion", "freebase_id": "/m/0755b"},
+    {"id": 375, "name": "Segway", "freebase_id": "/m/076bq"},
+    {"id": 376, "name": "Training bench", "freebase_id": "/m/076lb9"},
+    {"id": 377, "name": "Snake", "freebase_id": "/m/078jl"},
+    {"id": 378, "name": "Coffee table", "freebase_id": "/m/078n6m"},
+    {"id": 379, "name": "Skyscraper", "freebase_id": "/m/079cl"},
+    {"id": 380, "name": "Sheep", "freebase_id": "/m/07bgp"},
+    {"id": 381, "name": "Television", "freebase_id": "/m/07c52"},
+    {"id": 382, "name": "Trombone", "freebase_id": "/m/07c6l"},
+    {"id": 383, "name": "Tea", "freebase_id": "/m/07clx"},
+    {"id": 384, "name": "Tank", "freebase_id": "/m/07cmd"},
+    {"id": 385, "name": "Taco", "freebase_id": "/m/07crc"},
+    {"id": 386, "name": "Telephone", "freebase_id": "/m/07cx4"},
+    {"id": 387, "name": "Torch", "freebase_id": "/m/07dd4"},
+    {"id": 388, "name": "Tiger", "freebase_id": "/m/07dm6"},
+    {"id": 389, "name": "Strawberry", "freebase_id": "/m/07fbm7"},
+    {"id": 390, "name": "Trumpet", "freebase_id": "/m/07gql"},
+    {"id": 391, "name": "Tree", "freebase_id": "/m/07j7r"},
+    {"id": 392, "name": "Tomato", "freebase_id": "/m/07j87"},
+    {"id": 393, "name": "Train", "freebase_id": "/m/07jdr"},
+    {"id": 394, "name": "Tool", "freebase_id": "/m/07k1x"},
+    {"id": 395, "name": "Picnic basket", "freebase_id": "/m/07kng9"},
+    {"id": 396, "name": "Cooking spray", "freebase_id": "/m/07mcwg"},
+    {"id": 397, "name": "Trousers", "freebase_id": "/m/07mhn"},
+    {"id": 398, "name": "Bowling equipment", "freebase_id": "/m/07pj7bq"},
+    {"id": 399, "name": "Football helmet", "freebase_id": "/m/07qxg_"},
+    {"id": 400, "name": "Truck", "freebase_id": "/m/07r04"},
+    {"id": 401, "name": "Measuring cup", "freebase_id": "/m/07v9_z"},
+    {"id": 402, "name": "Coffeemaker", "freebase_id": "/m/07xyvk"},
+    {"id": 403, "name": "Violin", "freebase_id": "/m/07y_7"},
+    {"id": 404, "name": "Vehicle", "freebase_id": "/m/07yv9"},
+    {"id": 405, "name": "Handbag", "freebase_id": "/m/080hkjn"},
+    {"id": 406, "name": "Paper cutter", "freebase_id": "/m/080n7g"},
+    {"id": 407, "name": "Wine", "freebase_id": "/m/081qc"},
+    {"id": 408, "name": "Weapon", "freebase_id": "/m/083kb"},
+    {"id": 409, "name": "Wheel", "freebase_id": "/m/083wq"},
+    {"id": 410, "name": "Worm", "freebase_id": "/m/084hf"},
+    {"id": 411, "name": "Wok", "freebase_id": "/m/084rd"},
+    {"id": 412, "name": "Whale", "freebase_id": "/m/084zz"},
+    {"id": 413, "name": "Zebra", "freebase_id": "/m/0898b"},
+    {"id": 414, "name": "Auto part", "freebase_id": "/m/08dz3q"},
+    {"id": 415, "name": "Jug", "freebase_id": "/m/08hvt4"},
+    {"id": 416, "name": "Pizza cutter", "freebase_id": "/m/08ks85"},
+    {"id": 417, "name": "Cream", "freebase_id": "/m/08p92x"},
+    {"id": 418, "name": "Monkey", "freebase_id": "/m/08pbxl"},
+    {"id": 419, "name": "Lion", "freebase_id": "/m/096mb"},
+    {"id": 420, "name": "Bread", "freebase_id": "/m/09728"},
+    {"id": 421, "name": "Platter", "freebase_id": "/m/099ssp"},
+    {"id": 422, "name": "Chicken", "freebase_id": "/m/09b5t"},
+    {"id": 423, "name": "Eagle", "freebase_id": "/m/09csl"},
+    {"id": 424, "name": "Helicopter", "freebase_id": "/m/09ct_"},
+    {"id": 425, "name": "Owl", "freebase_id": "/m/09d5_"},
+    {"id": 426, "name": "Duck", "freebase_id": "/m/09ddx"},
+    {"id": 427, "name": "Turtle", "freebase_id": "/m/09dzg"},
+    {"id": 428, "name": "Hippopotamus", "freebase_id": "/m/09f20"},
+    {"id": 429, "name": "Crocodile", "freebase_id": "/m/09f_2"},
+    {"id": 430, "name": "Toilet", "freebase_id": "/m/09g1w"},
+    {"id": 431, "name": "Toilet paper", "freebase_id": "/m/09gtd"},
+    {"id": 432, "name": "Squid", "freebase_id": "/m/09gys"},
+    {"id": 433, "name": "Clothing", "freebase_id": "/m/09j2d"},
+    {"id": 434, "name": "Footwear", "freebase_id": "/m/09j5n"},
+    {"id": 435, "name": "Lemon", "freebase_id": "/m/09k_b"},
+    {"id": 436, "name": "Spider", "freebase_id": "/m/09kmb"},
+    {"id": 437, "name": "Deer", "freebase_id": "/m/09kx5"},
+    {"id": 438, "name": "Frog", "freebase_id": "/m/09ld4"},
+    {"id": 439, "name": "Banana", "freebase_id": "/m/09qck"},
+    {"id": 440, "name": "Rocket", "freebase_id": "/m/09rvcxw"},
+    {"id": 441, "name": "Wine glass", "freebase_id": "/m/09tvcd"},
+    {"id": 442, "name": "Countertop", "freebase_id": "/m/0b3fp9"},
+    {"id": 443, "name": "Tablet computer", "freebase_id": "/m/0bh9flk"},
+    {"id": 444, "name": "Waste container", "freebase_id": "/m/0bjyj5"},
+    {"id": 445, "name": "Swimming pool", "freebase_id": "/m/0b_rs"},
+    {"id": 446, "name": "Dog", "freebase_id": "/m/0bt9lr"},
+    {"id": 447, "name": "Book", "freebase_id": "/m/0bt_c3"},
+    {"id": 448, "name": "Elephant", "freebase_id": "/m/0bwd_0j"},
+    {"id": 449, "name": "Shark", "freebase_id": "/m/0by6g"},
+    {"id": 450, "name": "Candle", "freebase_id": "/m/0c06p"},
+    {"id": 451, "name": "Leopard", "freebase_id": "/m/0c29q"},
+    {"id": 452, "name": "Axe", "freebase_id": "/m/0c2jj"},
+    {"id": 453, "name": "Hand dryer", "freebase_id": "/m/0c3m8g"},
+    {"id": 454, "name": "Soap dispenser", "freebase_id": "/m/0c3mkw"},
+    {"id": 455, "name": "Porcupine", "freebase_id": "/m/0c568"},
+    {"id": 456, "name": "Flower", "freebase_id": "/m/0c9ph5"},
+    {"id": 457, "name": "Canary", "freebase_id": "/m/0ccs93"},
+    {"id": 458, "name": "Cheetah", "freebase_id": "/m/0cd4d"},
+    {"id": 459, "name": "Palm tree", "freebase_id": "/m/0cdl1"},
+    {"id": 460, "name": "Hamburger", "freebase_id": "/m/0cdn1"},
+    {"id": 461, "name": "Maple", "freebase_id": "/m/0cffdh"},
+    {"id": 462, "name": "Building", "freebase_id": "/m/0cgh4"},
+    {"id": 463, "name": "Fish", "freebase_id": "/m/0ch_cf"},
+    {"id": 464, "name": "Lobster", "freebase_id": "/m/0cjq5"},
+    {"id": 465, "name": "Garden Asparagus", "freebase_id": "/m/0cjs7"},
+    {"id": 466, "name": "Furniture", "freebase_id": "/m/0c_jw"},
+    {"id": 467, "name": "Hedgehog", "freebase_id": "/m/0cl4p"},
+    {"id": 468, "name": "Airplane", "freebase_id": "/m/0cmf2"},
+    {"id": 469, "name": "Spoon", "freebase_id": "/m/0cmx8"},
+    {"id": 470, "name": "Otter", "freebase_id": "/m/0cn6p"},
+    {"id": 471, "name": "Bull", "freebase_id": "/m/0cnyhnx"},
+    {"id": 472, "name": "Oyster", "freebase_id": "/m/0_cp5"},
+    {"id": 473, "name": "Horizontal bar", "freebase_id": "/m/0cqn2"},
+    {"id": 474, "name": "Convenience store", "freebase_id": "/m/0crjs"},
+    {"id": 475, "name": "Bomb", "freebase_id": "/m/0ct4f"},
+    {"id": 476, "name": "Bench", "freebase_id": "/m/0cvnqh"},
+    {"id": 477, "name": "Ice cream", "freebase_id": "/m/0cxn2"},
+    {"id": 478, "name": "Caterpillar", "freebase_id": "/m/0cydv"},
+    {"id": 479, "name": "Butterfly", "freebase_id": "/m/0cyf8"},
+    {"id": 480, "name": "Parachute", "freebase_id": "/m/0cyfs"},
+    {"id": 481, "name": "Orange", "freebase_id": "/m/0cyhj_"},
+    {"id": 482, "name": "Antelope", "freebase_id": "/m/0czz2"},
+    {"id": 483, "name": "Beaker", "freebase_id": "/m/0d20w4"},
+    {"id": 484, "name": "Moths and butterflies", "freebase_id": "/m/0d_2m"},
+    {"id": 485, "name": "Window", "freebase_id": "/m/0d4v4"},
+    {"id": 486, "name": "Closet", "freebase_id": "/m/0d4w1"},
+    {"id": 487, "name": "Castle", "freebase_id": "/m/0d5gx"},
+    {"id": 488, "name": "Jellyfish", "freebase_id": "/m/0d8zb"},
+    {"id": 489, "name": "Goose", "freebase_id": "/m/0dbvp"},
+    {"id": 490, "name": "Mule", "freebase_id": "/m/0dbzx"},
+    {"id": 491, "name": "Swan", "freebase_id": "/m/0dftk"},
+    {"id": 492, "name": "Peach", "freebase_id": "/m/0dj6p"},
+    {"id": 493, "name": "Coconut", "freebase_id": "/m/0djtd"},
+    {"id": 494, "name": "Seat belt", "freebase_id": "/m/0dkzw"},
+    {"id": 495, "name": "Raccoon", "freebase_id": "/m/0dq75"},
+    {"id": 496, "name": "Chisel", "freebase_id": "/m/0_dqb"},
+    {"id": 497, "name": "Fork", "freebase_id": "/m/0dt3t"},
+    {"id": 498, "name": "Lamp", "freebase_id": "/m/0dtln"},
+    {"id": 499, "name": "Camera", "freebase_id": "/m/0dv5r"},
+    {"id": 500, "name": "Squash (Plant)", "freebase_id": "/m/0dv77"},
+    {"id": 501, "name": "Racket", "freebase_id": "/m/0dv9c"},
+    {"id": 502, "name": "Human face", "freebase_id": "/m/0dzct"},
+    {"id": 503, "name": "Human arm", "freebase_id": "/m/0dzf4"},
+    {"id": 504, "name": "Vegetable", "freebase_id": "/m/0f4s2w"},
+    {"id": 505, "name": "Diaper", "freebase_id": "/m/0f571"},
+    {"id": 506, "name": "Unicycle", "freebase_id": "/m/0f6nr"},
+    {"id": 507, "name": "Falcon", "freebase_id": "/m/0f6wt"},
+    {"id": 508, "name": "Chime", "freebase_id": "/m/0f8s22"},
+    {"id": 509, "name": "Snail", "freebase_id": "/m/0f9_l"},
+    {"id": 510, "name": "Shellfish", "freebase_id": "/m/0fbdv"},
+    {"id": 511, "name": "Cabbage", "freebase_id": "/m/0fbw6"},
+    {"id": 512, "name": "Carrot", "freebase_id": "/m/0fj52s"},
+    {"id": 513, "name": "Mango", "freebase_id": "/m/0fldg"},
+    {"id": 514, "name": "Jeans", "freebase_id": "/m/0fly7"},
+    {"id": 515, "name": "Flowerpot", "freebase_id": "/m/0fm3zh"},
+    {"id": 516, "name": "Pineapple", "freebase_id": "/m/0fp6w"},
+    {"id": 517, "name": "Drawer", "freebase_id": "/m/0fqfqc"},
+    {"id": 518, "name": "Stool", "freebase_id": "/m/0fqt361"},
+    {"id": 519, "name": "Envelope", "freebase_id": "/m/0frqm"},
+    {"id": 520, "name": "Cake", "freebase_id": "/m/0fszt"},
+    {"id": 521, "name": "Dragonfly", "freebase_id": "/m/0ft9s"},
+    {"id": 522, "name": "Common sunflower", "freebase_id": "/m/0ftb8"},
+    {"id": 523, "name": "Microwave oven", "freebase_id": "/m/0fx9l"},
+    {"id": 524, "name": "Honeycomb", "freebase_id": "/m/0fz0h"},
+    {"id": 525, "name": "Marine mammal", "freebase_id": "/m/0gd2v"},
+    {"id": 526, "name": "Sea lion", "freebase_id": "/m/0gd36"},
+    {"id": 527, "name": "Ladybug", "freebase_id": "/m/0gj37"},
+    {"id": 528, "name": "Shelf", "freebase_id": "/m/0gjbg72"},
+    {"id": 529, "name": "Watch", "freebase_id": "/m/0gjkl"},
+    {"id": 530, "name": "Candy", "freebase_id": "/m/0gm28"},
+    {"id": 531, "name": "Salad", "freebase_id": "/m/0grw1"},
+    {"id": 532, "name": "Parrot", "freebase_id": "/m/0gv1x"},
+    {"id": 533, "name": "Handgun", "freebase_id": "/m/0gxl3"},
+    {"id": 534, "name": "Sparrow", "freebase_id": "/m/0h23m"},
+    {"id": 535, "name": "Van", "freebase_id": "/m/0h2r6"},
+    {"id": 536, "name": "Grinder", "freebase_id": "/m/0h8jyh6"},
+    {"id": 537, "name": "Spice rack", "freebase_id": "/m/0h8kx63"},
+    {"id": 538, "name": "Light bulb", "freebase_id": "/m/0h8l4fh"},
+    {"id": 539, "name": "Corded phone", "freebase_id": "/m/0h8lkj8"},
+    {"id": 540, "name": "Sports uniform", "freebase_id": "/m/0h8mhzd"},
+    {"id": 541, "name": "Tennis racket", "freebase_id": "/m/0h8my_4"},
+    {"id": 542, "name": "Wall clock", "freebase_id": "/m/0h8mzrc"},
+    {"id": 543, "name": "Serving tray", "freebase_id": "/m/0h8n27j"},
+    {"id": 544, "name": "Kitchen & dining room table", "freebase_id": "/m/0h8n5zk"},
+    {"id": 545, "name": "Dog bed", "freebase_id": "/m/0h8n6f9"},
+    {"id": 546, "name": "Cake stand", "freebase_id": "/m/0h8n6ft"},
+    {"id": 547, "name": "Cat furniture", "freebase_id": "/m/0h8nm9j"},
+    {"id": 548, "name": "Bathroom accessory", "freebase_id": "/m/0h8nr_l"},
+    {"id": 549, "name": "Facial tissue holder", "freebase_id": "/m/0h8nsvg"},
+    {"id": 550, "name": "Pressure cooker", "freebase_id": "/m/0h8ntjv"},
+    {"id": 551, "name": "Kitchen appliance", "freebase_id": "/m/0h99cwc"},
+    {"id": 552, "name": "Tire", "freebase_id": "/m/0h9mv"},
+    {"id": 553, "name": "Ruler", "freebase_id": "/m/0hdln"},
+    {"id": 554, "name": "Luggage and bags", "freebase_id": "/m/0hf58v5"},
+    {"id": 555, "name": "Microphone", "freebase_id": "/m/0hg7b"},
+    {"id": 556, "name": "Broccoli", "freebase_id": "/m/0hkxq"},
+    {"id": 557, "name": "Umbrella", "freebase_id": "/m/0hnnb"},
+    {"id": 558, "name": "Pastry", "freebase_id": "/m/0hnyx"},
+    {"id": 559, "name": "Grapefruit", "freebase_id": "/m/0hqkz"},
+    {"id": 560, "name": "Band-aid", "freebase_id": "/m/0j496"},
+    {"id": 561, "name": "Animal", "freebase_id": "/m/0jbk"},
+    {"id": 562, "name": "Bell pepper", "freebase_id": "/m/0jg57"},
+    {"id": 563, "name": "Turkey", "freebase_id": "/m/0jly1"},
+    {"id": 564, "name": "Lily", "freebase_id": "/m/0jqgx"},
+    {"id": 565, "name": "Pomegranate", "freebase_id": "/m/0jwn_"},
+    {"id": 566, "name": "Doughnut", "freebase_id": "/m/0jy4k"},
+    {"id": 567, "name": "Glasses", "freebase_id": "/m/0jyfg"},
+    {"id": 568, "name": "Human nose", "freebase_id": "/m/0k0pj"},
+    {"id": 569, "name": "Pen", "freebase_id": "/m/0k1tl"},
+    {"id": 570, "name": "Ant", "freebase_id": "/m/0_k2"},
+    {"id": 571, "name": "Car", "freebase_id": "/m/0k4j"},
+    {"id": 572, "name": "Aircraft", "freebase_id": "/m/0k5j"},
+    {"id": 573, "name": "Human hand", "freebase_id": "/m/0k65p"},
+    {"id": 574, "name": "Skunk", "freebase_id": "/m/0km7z"},
+    {"id": 575, "name": "Teddy bear", "freebase_id": "/m/0kmg4"},
+    {"id": 576, "name": "Watermelon", "freebase_id": "/m/0kpqd"},
+    {"id": 577, "name": "Cantaloupe", "freebase_id": "/m/0kpt_"},
+    {"id": 578, "name": "Dishwasher", "freebase_id": "/m/0ky7b"},
+    {"id": 579, "name": "Flute", "freebase_id": "/m/0l14j_"},
+    {"id": 580, "name": "Balance beam", "freebase_id": "/m/0l3ms"},
+    {"id": 581, "name": "Sandwich", "freebase_id": "/m/0l515"},
+    {"id": 582, "name": "Shrimp", "freebase_id": "/m/0ll1f78"},
+    {"id": 583, "name": "Sewing machine", "freebase_id": "/m/0llzx"},
+    {"id": 584, "name": "Binoculars", "freebase_id": "/m/0lt4_"},
+    {"id": 585, "name": "Rays and skates", "freebase_id": "/m/0m53l"},
+    {"id": 586, "name": "Ipod", "freebase_id": "/m/0mcx2"},
+    {"id": 587, "name": "Accordion", "freebase_id": "/m/0mkg"},
+    {"id": 588, "name": "Willow", "freebase_id": "/m/0mw_6"},
+    {"id": 589, "name": "Crab", "freebase_id": "/m/0n28_"},
+    {"id": 590, "name": "Crown", "freebase_id": "/m/0nl46"},
+    {"id": 591, "name": "Seahorse", "freebase_id": "/m/0nybt"},
+    {"id": 592, "name": "Perfume", "freebase_id": "/m/0p833"},
+    {"id": 593, "name": "Alpaca", "freebase_id": "/m/0pcr"},
+    {"id": 594, "name": "Taxi", "freebase_id": "/m/0pg52"},
+    {"id": 595, "name": "Canoe", "freebase_id": "/m/0ph39"},
+    {"id": 596, "name": "Remote control", "freebase_id": "/m/0qjjc"},
+    {"id": 597, "name": "Wheelchair", "freebase_id": "/m/0qmmr"},
+    {"id": 598, "name": "Rugby ball", "freebase_id": "/m/0wdt60w"},
+    {"id": 599, "name": "Armadillo", "freebase_id": "/m/0xfy"},
+    {"id": 600, "name": "Maracas", "freebase_id": "/m/0xzly"},
+    {"id": 601, "name": "Helmet", "freebase_id": "/m/0zvk5"},
+]
+
+categories_seg = [
+    {"id": 1, "name": "Screwdriver", "freebase_id": "/m/01bms0"},
+    {"id": 2, "name": "Light switch", "freebase_id": "/m/03jbxj"},
+    {"id": 3, "name": "Doughnut", "freebase_id": "/m/0jy4k"},
+    {"id": 4, "name": "Toilet paper", "freebase_id": "/m/09gtd"},
+    {"id": 5, "name": "Wrench", "freebase_id": "/m/01j5ks"},
+    {"id": 6, "name": "Toaster", "freebase_id": "/m/01k6s3"},
+    {"id": 7, "name": "Tennis ball", "freebase_id": "/m/05ctyq"},
+    {"id": 8, "name": "Radish", "freebase_id": "/m/015x5n"},
+    {"id": 9, "name": "Pomegranate", "freebase_id": "/m/0jwn_"},
+    {"id": 10, "name": "Kite", "freebase_id": "/m/02zt3"},
+    {"id": 11, "name": "Table tennis racket", "freebase_id": "/m/05_5p_0"},
+    {"id": 12, "name": "Hamster", "freebase_id": "/m/03qrc"},
+    {"id": 13, "name": "Barge", "freebase_id": "/m/01btn"},
+    {"id": 14, "name": "Shower", "freebase_id": "/m/02f9f_"},
+    {"id": 15, "name": "Printer", "freebase_id": "/m/01m4t"},
+    {"id": 16, "name": "Snowmobile", "freebase_id": "/m/01x3jk"},
+    {"id": 17, "name": "Fire hydrant", "freebase_id": "/m/01pns0"},
+    {"id": 18, "name": "Limousine", "freebase_id": "/m/01lcw4"},
+    {"id": 19, "name": "Whale", "freebase_id": "/m/084zz"},
+    {"id": 20, "name": "Microwave oven", "freebase_id": "/m/0fx9l"},
+    {"id": 21, "name": "Asparagus", "freebase_id": "/m/0cjs7"},
+    {"id": 22, "name": "Lion", "freebase_id": "/m/096mb"},
+    {"id": 23, "name": "Spatula", "freebase_id": "/m/02d1br"},
+    {"id": 24, "name": "Torch", "freebase_id": "/m/07dd4"},
+    {"id": 25, "name": "Volleyball", "freebase_id": "/m/02rgn06"},
+    {"id": 26, "name": "Ambulance", "freebase_id": "/m/012n7d"},
+    {"id": 27, "name": "Chopsticks", "freebase_id": "/m/01_5g"},
+    {"id": 28, "name": "Raccoon", "freebase_id": "/m/0dq75"},
+    {"id": 29, "name": "Blue jay", "freebase_id": "/m/01f8m5"},
+    {"id": 30, "name": "Lynx", "freebase_id": "/m/04g2r"},
+    {"id": 31, "name": "Dice", "freebase_id": "/m/029b3"},
+    {"id": 32, "name": "Filing cabinet", "freebase_id": "/m/047j0r"},
+    {"id": 33, "name": "Ruler", "freebase_id": "/m/0hdln"},
+    {"id": 34, "name": "Power plugs and sockets", "freebase_id": "/m/03bbps"},
+    {"id": 35, "name": "Bell pepper", "freebase_id": "/m/0jg57"},
+    {"id": 36, "name": "Binoculars", "freebase_id": "/m/0lt4_"},
+    {"id": 37, "name": "Pretzel", "freebase_id": "/m/01f91_"},
+    {"id": 38, "name": "Hot dog", "freebase_id": "/m/01b9xk"},
+    {"id": 39, "name": "Missile", "freebase_id": "/m/04ylt"},
+    {"id": 40, "name": "Common fig", "freebase_id": "/m/043nyj"},
+    {"id": 41, "name": "Croissant", "freebase_id": "/m/015wgc"},
+    {"id": 42, "name": "Adhesive tape", "freebase_id": "/m/03m3vtv"},
+    {"id": 43, "name": "Slow cooker", "freebase_id": "/m/02tsc9"},
+    {"id": 44, "name": "Dog bed", "freebase_id": "/m/0h8n6f9"},
+    {"id": 45, "name": "Harpsichord", "freebase_id": "/m/03q5t"},
+    {"id": 46, "name": "Billiard table", "freebase_id": "/m/04p0qw"},
+    {"id": 47, "name": "Alpaca", "freebase_id": "/m/0pcr"},
+    {"id": 48, "name": "Harbor seal", "freebase_id": "/m/02l8p9"},
+    {"id": 49, "name": "Grape", "freebase_id": "/m/0388q"},
+    {"id": 50, "name": "Nail", "freebase_id": "/m/05bm6"},
+    {"id": 51, "name": "Paper towel", "freebase_id": "/m/02w3r3"},
+    {"id": 52, "name": "Alarm clock", "freebase_id": "/m/046dlr"},
+    {"id": 53, "name": "Guacamole", "freebase_id": "/m/02g30s"},
+    {"id": 54, "name": "Starfish", "freebase_id": "/m/01h8tj"},
+    {"id": 55, "name": "Zebra", "freebase_id": "/m/0898b"},
+    {"id": 56, "name": "Segway", "freebase_id": "/m/076bq"},
+    {"id": 57, "name": "Sea turtle", "freebase_id": "/m/0120dh"},
+    {"id": 58, "name": "Scissors", "freebase_id": "/m/01lsmm"},
+    {"id": 59, "name": "Rhinoceros", "freebase_id": "/m/03d443"},
+    {"id": 60, "name": "Kangaroo", "freebase_id": "/m/04c0y"},
+    {"id": 61, "name": "Jaguar", "freebase_id": "/m/0449p"},
+    {"id": 62, "name": "Leopard", "freebase_id": "/m/0c29q"},
+    {"id": 63, "name": "Dumbbell", "freebase_id": "/m/04h8sr"},
+    {"id": 64, "name": "Envelope", "freebase_id": "/m/0frqm"},
+    {"id": 65, "name": "Winter melon", "freebase_id": "/m/02cvgx"},
+    {"id": 66, "name": "Teapot", "freebase_id": "/m/01fh4r"},
+    {"id": 67, "name": "Camel", "freebase_id": "/m/01x_v"},
+    {"id": 68, "name": "Beaker", "freebase_id": "/m/0d20w4"},
+    {"id": 69, "name": "Brown bear", "freebase_id": "/m/01dxs"},
+    {"id": 70, "name": "Toilet", "freebase_id": "/m/09g1w"},
+    {"id": 71, "name": "Teddy bear", "freebase_id": "/m/0kmg4"},
+    {"id": 72, "name": "Briefcase", "freebase_id": "/m/0584n8"},
+    {"id": 73, "name": "Stop sign", "freebase_id": "/m/02pv19"},
+    {"id": 74, "name": "Tiger", "freebase_id": "/m/07dm6"},
+    {"id": 75, "name": "Cabbage", "freebase_id": "/m/0fbw6"},
+    {"id": 76, "name": "Giraffe", "freebase_id": "/m/03bk1"},
+    {"id": 77, "name": "Polar bear", "freebase_id": "/m/0633h"},
+    {"id": 78, "name": "Shark", "freebase_id": "/m/0by6g"},
+    {"id": 79, "name": "Rabbit", "freebase_id": "/m/06mf6"},
+    {"id": 80, "name": "Swim cap", "freebase_id": "/m/04tn4x"},
+    {"id": 81, "name": "Pressure cooker", "freebase_id": "/m/0h8ntjv"},
+    {"id": 82, "name": "Kitchen knife", "freebase_id": "/m/058qzx"},
+    {"id": 83, "name": "Submarine sandwich", "freebase_id": "/m/06pcq"},
+    {"id": 84, "name": "Flashlight", "freebase_id": "/m/01kb5b"},
+    {"id": 85, "name": "Penguin", "freebase_id": "/m/05z6w"},
+    {"id": 86, "name": "Snake", "freebase_id": "/m/078jl"},
+    {"id": 87, "name": "Zucchini", "freebase_id": "/m/027pcv"},
+    {"id": 88, "name": "Bat", "freebase_id": "/m/01h44"},
+    {"id": 89, "name": "Food processor", "freebase_id": "/m/03y6mg"},
+    {"id": 90, "name": "Ostrich", "freebase_id": "/m/05n4y"},
+    {"id": 91, "name": "Sea lion", "freebase_id": "/m/0gd36"},
+    {"id": 92, "name": "Goldfish", "freebase_id": "/m/03fj2"},
+    {"id": 93, "name": "Elephant", "freebase_id": "/m/0bwd_0j"},
+    {"id": 94, "name": "Rocket", "freebase_id": "/m/09rvcxw"},
+    {"id": 95, "name": "Mouse", "freebase_id": "/m/04rmv"},
+    {"id": 96, "name": "Oyster", "freebase_id": "/m/0_cp5"},
+    {"id": 97, "name": "Digital clock", "freebase_id": "/m/06_72j"},
+    {"id": 98, "name": "Otter", "freebase_id": "/m/0cn6p"},
+    {"id": 99, "name": "Dolphin", "freebase_id": "/m/02hj4"},
+    {"id": 100, "name": "Punching bag", "freebase_id": "/m/0420v5"},
+    {"id": 101, "name": "Corded phone", "freebase_id": "/m/0h8lkj8"},
+    {"id": 102, "name": "Tennis racket", "freebase_id": "/m/0h8my_4"},
+    {"id": 103, "name": "Pancake", "freebase_id": "/m/01dwwc"},
+    {"id": 104, "name": "Mango", "freebase_id": "/m/0fldg"},
+    {"id": 105, "name": "Crocodile", "freebase_id": "/m/09f_2"},
+    {"id": 106, "name": "Waffle", "freebase_id": "/m/01dwsz"},
+    {"id": 107, "name": "Computer mouse", "freebase_id": "/m/020lf"},
+    {"id": 108, "name": "Kettle", "freebase_id": "/m/03s_tn"},
+    {"id": 109, "name": "Tart", "freebase_id": "/m/02zvsm"},
+    {"id": 110, "name": "Oven", "freebase_id": "/m/029bxz"},
+    {"id": 111, "name": "Banana", "freebase_id": "/m/09qck"},
+    {"id": 112, "name": "Cheetah", "freebase_id": "/m/0cd4d"},
+    {"id": 113, "name": "Raven", "freebase_id": "/m/06j2d"},
+    {"id": 114, "name": "Frying pan", "freebase_id": "/m/04v6l4"},
+    {"id": 115, "name": "Pear", "freebase_id": "/m/061_f"},
+    {"id": 116, "name": "Fox", "freebase_id": "/m/0306r"},
+    {"id": 117, "name": "Skateboard", "freebase_id": "/m/06_fw"},
+    {"id": 118, "name": "Rugby ball", "freebase_id": "/m/0wdt60w"},
+    {"id": 119, "name": "Watermelon", "freebase_id": "/m/0kpqd"},
+    {"id": 120, "name": "Flute", "freebase_id": "/m/0l14j_"},
+    {"id": 121, "name": "Canary", "freebase_id": "/m/0ccs93"},
+    {"id": 122, "name": "Door handle", "freebase_id": "/m/03c7gz"},
+    {"id": 123, "name": "Saxophone", "freebase_id": "/m/06ncr"},
+    {"id": 124, "name": "Burrito", "freebase_id": "/m/01j3zr"},
+    {"id": 125, "name": "Suitcase", "freebase_id": "/m/01s55n"},
+    {"id": 126, "name": "Roller skates", "freebase_id": "/m/02p3w7d"},
+    {"id": 127, "name": "Dagger", "freebase_id": "/m/02gzp"},
+    {"id": 128, "name": "Seat belt", "freebase_id": "/m/0dkzw"},
+    {"id": 129, "name": "Washing machine", "freebase_id": "/m/0174k2"},
+    {"id": 130, "name": "Jet ski", "freebase_id": "/m/01xs3r"},
+    {"id": 131, "name": "Sombrero", "freebase_id": "/m/02jfl0"},
+    {"id": 132, "name": "Pig", "freebase_id": "/m/068zj"},
+    {"id": 133, "name": "Drinking straw", "freebase_id": "/m/03v5tg"},
+    {"id": 134, "name": "Peach", "freebase_id": "/m/0dj6p"},
+    {"id": 135, "name": "Tortoise", "freebase_id": "/m/011k07"},
+    {"id": 136, "name": "Towel", "freebase_id": "/m/0162_1"},
+    {"id": 137, "name": "Tablet computer", "freebase_id": "/m/0bh9flk"},
+    {"id": 138, "name": "Cucumber", "freebase_id": "/m/015x4r"},
+    {"id": 139, "name": "Mule", "freebase_id": "/m/0dbzx"},
+    {"id": 140, "name": "Potato", "freebase_id": "/m/05vtc"},
+    {"id": 141, "name": "Frog", "freebase_id": "/m/09ld4"},
+    {"id": 142, "name": "Bear", "freebase_id": "/m/01dws"},
+    {"id": 143, "name": "Lighthouse", "freebase_id": "/m/04h7h"},
+    {"id": 144, "name": "Belt", "freebase_id": "/m/0176mf"},
+    {"id": 145, "name": "Baseball bat", "freebase_id": "/m/03g8mr"},
+    {"id": 146, "name": "Racket", "freebase_id": "/m/0dv9c"},
+    {"id": 147, "name": "Sword", "freebase_id": "/m/06y5r"},
+    {"id": 148, "name": "Bagel", "freebase_id": "/m/01fb_0"},
+    {"id": 149, "name": "Goat", "freebase_id": "/m/03fwl"},
+    {"id": 150, "name": "Lizard", "freebase_id": "/m/04m9y"},
+    {"id": 151, "name": "Parrot", "freebase_id": "/m/0gv1x"},
+    {"id": 152, "name": "Owl", "freebase_id": "/m/09d5_"},
+    {"id": 153, "name": "Turkey", "freebase_id": "/m/0jly1"},
+    {"id": 154, "name": "Cello", "freebase_id": "/m/01xqw"},
+    {"id": 155, "name": "Knife", "freebase_id": "/m/04ctx"},
+    {"id": 156, "name": "Handgun", "freebase_id": "/m/0gxl3"},
+    {"id": 157, "name": "Carrot", "freebase_id": "/m/0fj52s"},
+    {"id": 158, "name": "Hamburger", "freebase_id": "/m/0cdn1"},
+    {"id": 159, "name": "Grapefruit", "freebase_id": "/m/0hqkz"},
+    {"id": 160, "name": "Tap", "freebase_id": "/m/02jz0l"},
+    {"id": 161, "name": "Tea", "freebase_id": "/m/07clx"},
+    {"id": 162, "name": "Bull", "freebase_id": "/m/0cnyhnx"},
+    {"id": 163, "name": "Turtle", "freebase_id": "/m/09dzg"},
+    {"id": 164, "name": "Bust", "freebase_id": "/m/04yqq2"},
+    {"id": 165, "name": "Monkey", "freebase_id": "/m/08pbxl"},
+    {"id": 166, "name": "Wok", "freebase_id": "/m/084rd"},
+    {"id": 167, "name": "Broccoli", "freebase_id": "/m/0hkxq"},
+    {"id": 168, "name": "Pitcher", "freebase_id": "/m/054fyh"},
+    {"id": 169, "name": "Whiteboard", "freebase_id": "/m/02d9qx"},
+    {"id": 170, "name": "Squirrel", "freebase_id": "/m/071qp"},
+    {"id": 171, "name": "Jug", "freebase_id": "/m/08hvt4"},
+    {"id": 172, "name": "Woodpecker", "freebase_id": "/m/01dy8n"},
+    {"id": 173, "name": "Pizza", "freebase_id": "/m/0663v"},
+    {"id": 174, "name": "Surfboard", "freebase_id": "/m/019w40"},
+    {"id": 175, "name": "Sofa bed", "freebase_id": "/m/03m3pdh"},
+    {"id": 176, "name": "Sheep", "freebase_id": "/m/07bgp"},
+    {"id": 177, "name": "Candle", "freebase_id": "/m/0c06p"},
+    {"id": 178, "name": "Muffin", "freebase_id": "/m/01tcjp"},
+    {"id": 179, "name": "Cookie", "freebase_id": "/m/021mn"},
+    {"id": 180, "name": "Apple", "freebase_id": "/m/014j1m"},
+    {"id": 181, "name": "Chest of drawers", "freebase_id": "/m/05kyg_"},
+    {"id": 182, "name": "Skull", "freebase_id": "/m/016m2d"},
+    {"id": 183, "name": "Chicken", "freebase_id": "/m/09b5t"},
+    {"id": 184, "name": "Loveseat", "freebase_id": "/m/0703r8"},
+    {"id": 185, "name": "Baseball glove", "freebase_id": "/m/03grzl"},
+    {"id": 186, "name": "Piano", "freebase_id": "/m/05r5c"},
+    {"id": 187, "name": "Waste container", "freebase_id": "/m/0bjyj5"},
+    {"id": 188, "name": "Barrel", "freebase_id": "/m/02zn6n"},
+    {"id": 189, "name": "Swan", "freebase_id": "/m/0dftk"},
+    {"id": 190, "name": "Taxi", "freebase_id": "/m/0pg52"},
+    {"id": 191, "name": "Lemon", "freebase_id": "/m/09k_b"},
+    {"id": 192, "name": "Pumpkin", "freebase_id": "/m/05zsy"},
+    {"id": 193, "name": "Sparrow", "freebase_id": "/m/0h23m"},
+    {"id": 194, "name": "Orange", "freebase_id": "/m/0cyhj_"},
+    {"id": 195, "name": "Tank", "freebase_id": "/m/07cmd"},
+    {"id": 196, "name": "Sandwich", "freebase_id": "/m/0l515"},
+    {"id": 197, "name": "Coffee", "freebase_id": "/m/02vqfm"},
+    {"id": 198, "name": "Juice", "freebase_id": "/m/01z1kdw"},
+    {"id": 199, "name": "Coin", "freebase_id": "/m/0242l"},
+    {"id": 200, "name": "Pen", "freebase_id": "/m/0k1tl"},
+    {"id": 201, "name": "Watch", "freebase_id": "/m/0gjkl"},
+    {"id": 202, "name": "Eagle", "freebase_id": "/m/09csl"},
+    {"id": 203, "name": "Goose", "freebase_id": "/m/0dbvp"},
+    {"id": 204, "name": "Falcon", "freebase_id": "/m/0f6wt"},
+    {"id": 205, "name": "Christmas tree", "freebase_id": "/m/025nd"},
+    {"id": 206, "name": "Sunflower", "freebase_id": "/m/0ftb8"},
+    {"id": 207, "name": "Vase", "freebase_id": "/m/02s195"},
+    {"id": 208, "name": "Football", "freebase_id": "/m/01226z"},
+    {"id": 209, "name": "Canoe", "freebase_id": "/m/0ph39"},
+    {"id": 210, "name": "High heels", "freebase_id": "/m/06k2mb"},
+    {"id": 211, "name": "Spoon", "freebase_id": "/m/0cmx8"},
+    {"id": 212, "name": "Mug", "freebase_id": "/m/02jvh9"},
+    {"id": 213, "name": "Swimwear", "freebase_id": "/m/01gkx_"},
+    {"id": 214, "name": "Duck", "freebase_id": "/m/09ddx"},
+    {"id": 215, "name": "Cat", "freebase_id": "/m/01yrx"},
+    {"id": 216, "name": "Tomato", "freebase_id": "/m/07j87"},
+    {"id": 217, "name": "Cocktail", "freebase_id": "/m/024g6"},
+    {"id": 218, "name": "Clock", "freebase_id": "/m/01x3z"},
+    {"id": 219, "name": "Cowboy hat", "freebase_id": "/m/025rp__"},
+    {"id": 220, "name": "Miniskirt", "freebase_id": "/m/01cmb2"},
+    {"id": 221, "name": "Cattle", "freebase_id": "/m/01xq0k1"},
+    {"id": 222, "name": "Strawberry", "freebase_id": "/m/07fbm7"},
+    {"id": 223, "name": "Bronze sculpture", "freebase_id": "/m/01yx86"},
+    {"id": 224, "name": "Pillow", "freebase_id": "/m/034c16"},
+    {"id": 225, "name": "Squash", "freebase_id": "/m/0dv77"},
+    {"id": 226, "name": "Traffic light", "freebase_id": "/m/015qff"},
+    {"id": 227, "name": "Saucer", "freebase_id": "/m/03q5c7"},
+    {"id": 228, "name": "Reptile", "freebase_id": "/m/06bt6"},
+    {"id": 229, "name": "Cake", "freebase_id": "/m/0fszt"},
+    {"id": 230, "name": "Plastic bag", "freebase_id": "/m/05gqfk"},
+    {"id": 231, "name": "Studio couch", "freebase_id": "/m/026qbn5"},
+    {"id": 232, "name": "Beer", "freebase_id": "/m/01599"},
+    {"id": 233, "name": "Scarf", "freebase_id": "/m/02h19r"},
+    {"id": 234, "name": "Coffee cup", "freebase_id": "/m/02p5f1q"},
+    {"id": 235, "name": "Wine", "freebase_id": "/m/081qc"},
+    {"id": 236, "name": "Mushroom", "freebase_id": "/m/052sf"},
+    {"id": 237, "name": "Traffic sign", "freebase_id": "/m/01mqdt"},
+    {"id": 238, "name": "Camera", "freebase_id": "/m/0dv5r"},
+    {"id": 239, "name": "Rose", "freebase_id": "/m/06m11"},
+    {"id": 240, "name": "Couch", "freebase_id": "/m/02crq1"},
+    {"id": 241, "name": "Handbag", "freebase_id": "/m/080hkjn"},
+    {"id": 242, "name": "Fedora", "freebase_id": "/m/02fq_6"},
+    {"id": 243, "name": "Sock", "freebase_id": "/m/01nq26"},
+    {"id": 244, "name": "Computer keyboard", "freebase_id": "/m/01m2v"},
+    {"id": 245, "name": "Mobile phone", "freebase_id": "/m/050k8"},
+    {"id": 246, "name": "Ball", "freebase_id": "/m/018xm"},
+    {"id": 247, "name": "Balloon", "freebase_id": "/m/01j51"},
+    {"id": 248, "name": "Horse", "freebase_id": "/m/03k3r"},
+    {"id": 249, "name": "Boot", "freebase_id": "/m/01b638"},
+    {"id": 250, "name": "Fish", "freebase_id": "/m/0ch_cf"},
+    {"id": 251, "name": "Backpack", "freebase_id": "/m/01940j"},
+    {"id": 252, "name": "Skirt", "freebase_id": "/m/02wv6h6"},
+    {"id": 253, "name": "Van", "freebase_id": "/m/0h2r6"},
+    {"id": 254, "name": "Bread", "freebase_id": "/m/09728"},
+    {"id": 255, "name": "Glove", "freebase_id": "/m/0174n1"},
+    {"id": 256, "name": "Dog", "freebase_id": "/m/0bt9lr"},
+    {"id": 257, "name": "Airplane", "freebase_id": "/m/0cmf2"},
+    {"id": 258, "name": "Motorcycle", "freebase_id": "/m/04_sv"},
+    {"id": 259, "name": "Drink", "freebase_id": "/m/0271t"},
+    {"id": 260, "name": "Book", "freebase_id": "/m/0bt_c3"},
+    {"id": 261, "name": "Train", "freebase_id": "/m/07jdr"},
+    {"id": 262, "name": "Flower", "freebase_id": "/m/0c9ph5"},
+    {"id": 263, "name": "Carnivore", "freebase_id": "/m/01lrl"},
+    {"id": 264, "name": "Human ear", "freebase_id": "/m/039xj_"},
+    {"id": 265, "name": "Toy", "freebase_id": "/m/0138tl"},
+    {"id": 266, "name": "Box", "freebase_id": "/m/025dyy"},
+    {"id": 267, "name": "Truck", "freebase_id": "/m/07r04"},
+    {"id": 268, "name": "Wheel", "freebase_id": "/m/083wq"},
+    {"id": 269, "name": "Aircraft", "freebase_id": "/m/0k5j"},
+    {"id": 270, "name": "Bus", "freebase_id": "/m/01bjv"},
+    {"id": 271, "name": "Human mouth", "freebase_id": "/m/0283dt1"},
+    {"id": 272, "name": "Sculpture", "freebase_id": "/m/06msq"},
+    {"id": 273, "name": "Shirt", "freebase_id": "/m/01n4qj"},
+    {"id": 274, "name": "Hat", "freebase_id": "/m/02dl1y"},
+    {"id": 275, "name": "Vehicle registration plate", "freebase_id": "/m/01jfm_"},
+    {"id": 276, "name": "Guitar", "freebase_id": "/m/0342h"},
+    {"id": 277, "name": "Sun hat", "freebase_id": "/m/02wbtzl"},
+    {"id": 278, "name": "Bottle", "freebase_id": "/m/04dr76w"},
+    {"id": 279, "name": "Luggage and bags", "freebase_id": "/m/0hf58v5"},
+    {"id": 280, "name": "Trousers", "freebase_id": "/m/07mhn"},
+    {"id": 281, "name": "Bicycle wheel", "freebase_id": "/m/01bqk0"},
+    {"id": 282, "name": "Suit", "freebase_id": "/m/01xyhv"},
+    {"id": 283, "name": "Bowl", "freebase_id": "/m/04kkgm"},
+    {"id": 284, "name": "Man", "freebase_id": "/m/04yx4"},
+    {"id": 285, "name": "Flowerpot", "freebase_id": "/m/0fm3zh"},
+    {"id": 286, "name": "Laptop", "freebase_id": "/m/01c648"},
+    {"id": 287, "name": "Boy", "freebase_id": "/m/01bl7v"},
+    {"id": 288, "name": "Picture frame", "freebase_id": "/m/06z37_"},
+    {"id": 289, "name": "Bird", "freebase_id": "/m/015p6"},
+    {"id": 290, "name": "Car", "freebase_id": "/m/0k4j"},
+    {"id": 291, "name": "Shorts", "freebase_id": "/m/01bfm9"},
+    {"id": 292, "name": "Woman", "freebase_id": "/m/03bt1vf"},
+    {"id": 293, "name": "Platter", "freebase_id": "/m/099ssp"},
+    {"id": 294, "name": "Tie", "freebase_id": "/m/01rkbr"},
+    {"id": 295, "name": "Girl", "freebase_id": "/m/05r655"},
+    {"id": 296, "name": "Skyscraper", "freebase_id": "/m/079cl"},
+    {"id": 297, "name": "Person", "freebase_id": "/m/01g317"},
+    {"id": 298, "name": "Flag", "freebase_id": "/m/03120"},
+    {"id": 299, "name": "Jeans", "freebase_id": "/m/0fly7"},
+    {"id": 300, "name": "Dress", "freebase_id": "/m/01d40f"},
+]
+
+
+def _get_builtin_metadata(cats, class_image_count=None):
+    id_to_name = {x["id"]: x["name"] for x in cats}
+    thing_dataset_id_to_contiguous_id = {i + 1: i for i in range(len(cats))}
+    thing_classes = [x["name"] for x in sorted(cats, key=lambda x: x["id"])]
+    return {
+        "thing_dataset_id_to_contiguous_id": thing_dataset_id_to_contiguous_id,
+        "thing_classes": thing_classes,
+        "class_image_count": class_image_count,
+    }
+
+
+_PREDEFINED_SPLITS_OID = {
+    "oid_train": ("oid/images/train/", "oid/annotations/oid_challenge_2019_train_bbox.json"),
+    "oid_val": ("oid/images/validation/", "oid/annotations/oid_challenge_2019_val.json"),
+    "oid_val_expanded": (
+        "oid/images/validation/",
+        "oid/annotations/oid_challenge_2019_val_expanded.json",
+    ),
+    "oid_kaggle_test": ("oid/images/test/", "oid/annotations/oid_kaggle_test_image_info.json"),
+}
+
+
+for key, (image_root, json_file) in _PREDEFINED_SPLITS_OID.items():
+    register_oid_instances(
+        key,
+        _get_builtin_metadata(OPENIMAGES_2019_CATEGORIES),
+        os.path.join("datasets", json_file) if "://" not in json_file else json_file,
+        os.path.join("datasets", image_root),
+    )
+
+_PREDEFINED_SPLITS_OID_SEG = {
+    "oid_seg_train": ("oid/images/train/", "oid/annotations/openimages_instances_train.json"),
+    "oid_seg_val": ("oid/images/validation/", "oid/annotations/openimages_instances_val.json"),
+    "oid_seg_kaggle_test": (
+        "oid/images/test/",
+        "oid/annotations/openimages_instances_kaggle_test_image_info.json",
+    ),
+}
+
+
+for key, (image_root, json_file) in _PREDEFINED_SPLITS_OID_SEG.items():
+    register_oid_instances(
+        key,
+        _get_builtin_metadata(categories_seg),
+        os.path.join("datasets", json_file) if "://" not in json_file else json_file,
+        os.path.join("datasets", image_root),
+    )
+
+
+_PREDEFINED_SPLITS_OPENIMAGES_DETECTION = {
+    "openimages_challenge_2019_train": (
+        "openimages/train/",
+        "openimages/annotations/openimages_challenge_2019_train_bbox.json",
+    ),
+    "openimages_challenge_2019_val": (
+        "openimages/validation/",
+        "openimages/annotations/openimages_challenge_2019_val_bbox.json",
+    ),
+}
+
+
+for key, (image_root, json_file) in _PREDEFINED_SPLITS_OPENIMAGES_DETECTION.items():
+    register_oid_instances(
+        key,
+        _get_builtin_metadata(OPENIMAGES_2019_CATEGORIES),
+        os.path.join("datasets", json_file) if "://" not in json_file else json_file,
+        os.path.join("datasets", image_root),
+    )
+
+
+_PREDEFINED_SPLITS_OPENIMAGES_V6_DETECTION = {
+    "openimages_v6_train_bbox": (
+        "openimages/train/",
+        "openimages/annotations/openimages_v6_train_bbox.json",
+    ),
+    "openimages_v6_train_bbox_nogroup": (
+        "openimages/train/",
+        "openimages/annotations/openimages_v6_train_bbox_nogroup.json",
+    ),
+    "openimages_v6_val_bbox": (
+        "openimages/validation/",
+        "openimages/annotations/openimages_v6_val_bbox.json",
+    ),
+    "openimages_v6_val_bbox_nogroup": (
+        "openimages/validation/",
+        "openimages/annotations/openimages_v6_val_bbox_nogroup.json",
+    ),
+    "openimages_v6_train_instance": (
+        "openimages/train/",
+        "openimages/annotations/openimages_v6_train_instance.json",
+    ),
+    "openimages_v6_val_instance": (
+        "openimages/validation/",
+        "openimages/annotations/openimages_v6_val_instance.json",
+    ),
+    "openimages_v6_train_bbox_instance": (
+        "openimages/train/",
+        "openimages/annotations/openimages_v6_train_bbox_instance.json",
+    ),
+    "openimages_v6_val_bbox_instance": (
+        "openimages/validation/",
+        "openimages/annotations/openimages_v6_val_bbox_instance.json",
+    ),
+}
+
+
+def register_all_oid(root):
+    for key, (image_root, json_file) in _PREDEFINED_SPLITS_OPENIMAGES_V6_DETECTION.items():
+        register_oid_instances(
+            key,
+            _get_builtin_metadata(OPENIMAGES_V6_CATEGORIES, OPENIMAGES_v6_CATEGORY_IMAGE_COUNT),
+            os.path.join(root, json_file) if "://" not in json_file else json_file,
+            os.path.join(root, image_root),
+        )
+
+
+if __name__.endswith(".oid"):
+    # Assume pre-defined datasets live in `./datasets`.
+    _root = os.getenv("DETECTRON2_DATASETS", "datasets")
+    register_all_oid(_root)

approach/ovod/APE/ape/data/datasets/openimages_v6_category_image_count.py

@@ -0,0 +1,2 @@
+OPENIMAGES_v6_CATEGORY_IMAGE_COUNT = [{"id": 1, "name": "Tortoise", "freebase_id": "/m/011k07", "image_count": 1151, "instance_count": 1678}, {"id": 2, "name": "Container", "freebase_id": "/m/011q46kg", "image_count": 0, "instance_count": 0}, {"id": 3, "name": "Magpie", "freebase_id": "/m/012074", "image_count": 100, "instance_count": 117}, {"id": 4, "name": "Sea turtle", "freebase_id": "/m/0120dh", "image_count": 664, "instance_count": 999}, {"id": 5, "name": "Football", "freebase_id": "/m/01226z", "image_count": 2727, "instance_count": 3150}, {"id": 6, "name": "Ambulance", "freebase_id": "/m/012n7d", "image_count": 305, "instance_count": 391}, {"id": 7, "name": "Ladder", "freebase_id": "/m/012w5l", "image_count": 679, "instance_count": 895}, {"id": 8, "name": "Toothbrush", "freebase_id": "/m/012xff", "image_count": 106, "instance_count": 202}, {"id": 9, "name": "Syringe", "freebase_id": "/m/012ysf", "image_count": 92, "instance_count": 124}, {"id": 10, "name": "Sink", "freebase_id": "/m/0130jx", "image_count": 1165, "instance_count": 1327}, {"id": 11, "name": "Toy", "freebase_id": "/m/0138tl", "image_count": 13584, "instance_count": 43916}, {"id": 12, "name": "Organ (Musical Instrument)", "freebase_id": "/m/013y1f", "image_count": 366, "instance_count": 386}, {"id": 13, "name": "Cassette deck", "freebase_id": "/m/01432t", "image_count": 49, "instance_count": 66}, {"id": 14, "name": "Apple", "freebase_id": "/m/014j1m", "image_count": 630, "instance_count": 1624}, {"id": 15, "name": "Human eye", "freebase_id": "/m/014sv8", "image_count": 20295, "instance_count": 58440}, {"id": 16, "name": "Cosmetics", "freebase_id": "/m/014trl", "image_count": 790, "instance_count": 2090}, {"id": 17, "name": "Paddle", "freebase_id": "/m/014y4n", "image_count": 1705, "instance_count": 4598}, {"id": 18, "name": "Snowman", "freebase_id": "/m/0152hh", "image_count": 497, "instance_count": 696}, {"id": 19, "name": "Beer", "freebase_id": "/m/01599", "image_count": 4063, "instance_count": 7841}, {"id": 20, "name": "Chopsticks", "freebase_id": "/m/01_5g", "image_count": 353, "instance_count": 448}, {"id": 21, "name": "Human beard", "freebase_id": "/m/015h_t", "image_count": 2530, "instance_count": 2845}, {"id": 22, "name": "Bird", "freebase_id": "/m/015p6", "image_count": 15042, "instance_count": 32316}, {"id": 23, "name": "Parking meter", "freebase_id": "/m/015qbp", "image_count": 176, "instance_count": 208}, {"id": 24, "name": "Traffic light", "freebase_id": "/m/015qff", "image_count": 1188, "instance_count": 4663}, {"id": 25, "name": "Croissant", "freebase_id": "/m/015wgc", "image_count": 167, "instance_count": 307}, {"id": 26, "name": "Cucumber", "freebase_id": "/m/015x4r", "image_count": 200, "instance_count": 477}, {"id": 27, "name": "Radish", "freebase_id": "/m/015x5n", "image_count": 80, "instance_count": 325}, {"id": 28, "name": "Towel", "freebase_id": "/m/0162_1", "image_count": 118, "instance_count": 202}, {"id": 29, "name": "Doll", "freebase_id": "/m/0167gd", "image_count": 3049, "instance_count": 5446}, {"id": 30, "name": "Skull", "freebase_id": "/m/016m2d", "image_count": 1407, "instance_count": 2340}, {"id": 31, "name": "Washing machine", "freebase_id": "/m/0174k2", "image_count": 299, "instance_count": 593}, {"id": 32, "name": "Glove", "freebase_id": "/m/0174n1", "image_count": 656, "instance_count": 1071}, {"id": 33, "name": "Tick", "freebase_id": "/m/0175cv", "image_count": 95, "instance_count": 127}, {"id": 34, "name": "Belt", "freebase_id": "/m/0176mf", "image_count": 341, "instance_count": 411}, {"id": 35, "name": "Sunglasses", "freebase_id": "/m/017ftj", "image_count": 12875, "instance_count": 19324}, {"id": 36, "name": "Banjo", "freebase_id": "/m/018j2", "image_count": 235, "instance_count": 253}, {"id": 37, "name": "Cart", "freebase_id": "/m/018p4k", "image_count": 1376, "instance_count": 1910}, {"id": 38, "name": "Ball", "freebase_id": "/m/018xm", "image_count": 2479, "instance_count": 4943}, {"id": 39, "name": "Backpack", "freebase_id": "/m/01940j", "image_count": 598, "instance_count": 966}, {"id": 40, "name": "Bicycle", "freebase_id": "/m/0199g", "image_count": 11536, "instance_count": 21840}, {"id": 41, "name": "Home appliance", "freebase_id": "/m/019dx1", "image_count": 893, "instance_count": 1732}, {"id": 42, "name": "Centipede", "freebase_id": "/m/019h78", "image_count": 231, "instance_count": 275}, {"id": 43, "name": "Boat", "freebase_id": "/m/019jd", "image_count": 19756, "instance_count": 54017}, {"id": 44, "name": "Surfboard", "freebase_id": "/m/019w40", "image_count": 1641, "instance_count": 2427}, {"id": 45, "name": "Boot", "freebase_id": "/m/01b638", "image_count": 1311, "instance_count": 2812}, {"id": 46, "name": "Headphones", "freebase_id": "/m/01b7fy", "image_count": 1028, "instance_count": 1193}, {"id": 47, "name": "Hot dog", "freebase_id": "/m/01b9xk", "image_count": 311, "instance_count": 405}, {"id": 48, "name": "Shorts", "freebase_id": "/m/01bfm9", "image_count": 5427, "instance_count": 12674}, {"id": 49, "name": "Fast food", "freebase_id": "/m/01_bhs", "image_count": 4621, "instance_count": 11599}, {"id": 50, "name": "Bus", "freebase_id": "/m/01bjv", "image_count": 5811, "instance_count": 9584}, {"id": 51, "name": "Boy", "freebase_id": "/m/01bl7v", "image_count": 36079, "instance_count": 61661}, {"id": 52, "name": "Screwdriver", "freebase_id": "/m/01bms0", "image_count": 46, "instance_count": 85}, {"id": 53, "name": "Bicycle wheel", "freebase_id": "/m/01bqk0", "image_count": 10655, "instance_count": 31513}, {"id": 54, "name": "Barge", "freebase_id": "/m/01btn", "image_count": 295, "instance_count": 690}, {"id": 55, "name": "Laptop", "freebase_id": "/m/01c648", "image_count": 5171, "instance_count": 8213}, {"id": 56, "name": "Miniskirt", "freebase_id": "/m/01cmb2", "image_count": 674, "instance_count": 851}, {"id": 57, "name": "Drill (Tool)", "freebase_id": "/m/01d380", "image_count": 169, "instance_count": 201}, {"id": 58, "name": "Dress", "freebase_id": "/m/01d40f", "image_count": 25529, "instance_count": 41137}, {"id": 59, "name": "Bear", "freebase_id": "/m/01dws", "image_count": 270, "instance_count": 326}, {"id": 60, "name": "Waffle", "freebase_id": "/m/01dwsz", "image_count": 327, "instance_count": 581}, {"id": 61, "name": "Pancake", "freebase_id": "/m/01dwwc", "image_count": 181, "instance_count": 437}, {"id": 62, "name": "Brown bear", "freebase_id": "/m/01dxs", "image_count": 403, "instance_count": 532}, {"id": 63, "name": "Woodpecker", "freebase_id": "/m/01dy8n", "image_count": 391, "instance_count": 480}, {"id": 64, "name": "Blue jay", "freebase_id": "/m/01f8m5", "image_count": 210, "instance_count": 233}, {"id": 65, "name": "Pretzel", "freebase_id": "/m/01f91_", "image_count": 123, "instance_count": 220}, {"id": 66, "name": "Bagel", "freebase_id": "/m/01fb_0", "image_count": 173, "instance_count": 374}, {"id": 67, "name": "Tower", "freebase_id": "/m/01fdzj", "image_count": 16337, "instance_count": 44982}, {"id": 68, "name": "Teapot", "freebase_id": "/m/01fh4r", "image_count": 448, "instance_count": 564}, {"id": 69, "name": "Person", "freebase_id": "/m/01g317", "image_count": 177356, "instance_count": 595849}, {"id": 70, "name": "Bow and arrow", "freebase_id": "/m/01g3x7", "image_count": 235, "instance_count": 459}, {"id": 71, "name": "Swimwear", "freebase_id": "/m/01gkx_", "image_count": 3084, "instance_count": 8645}, {"id": 72, "name": "Beehive", "freebase_id": "/m/01gllr", "image_count": 283, "instance_count": 327}, {"id": 73, "name": "Brassiere", "freebase_id": "/m/01gmv2", "image_count": 1146, "instance_count": 1567}, {"id": 74, "name": "Bee", "freebase_id": "/m/01h3n", "image_count": 2113, "instance_count": 4484}, {"id": 75, "name": "Bat (Animal)", "freebase_id": "/m/01h44", "image_count": 331, "instance_count": 490}, {"id": 76, "name": "Starfish", "freebase_id": "/m/01h8tj", "image_count": 409, "instance_count": 577}, {"id": 77, "name": "Popcorn", "freebase_id": "/m/01hrv5", "image_count": 179, "instance_count": 245}, {"id": 78, "name": "Burrito", "freebase_id": "/m/01j3zr", "image_count": 160, "instance_count": 251}, {"id": 79, "name": "Chainsaw", "freebase_id": "/m/01j4z9", "image_count": 100, "instance_count": 106}, {"id": 80, "name": "Balloon", "freebase_id": "/m/01j51", "image_count": 1691, "instance_count": 8440}, {"id": 81, "name": "Wrench", "freebase_id": "/m/01j5ks", "image_count": 54, "instance_count": 134}, {"id": 82, "name": "Tent", "freebase_id": "/m/01j61q", "image_count": 1588, "instance_count": 2929}, {"id": 83, "name": "Vehicle registration plate", "freebase_id": "/m/01jfm_", "image_count": 3956, "instance_count": 5420}, {"id": 84, "name": "Lantern", "freebase_id": "/m/01jfsr", "image_count": 555, "instance_count": 2221}, {"id": 85, "name": "Toaster", "freebase_id": "/m/01k6s3", "image_count": 56, "instance_count": 68}, {"id": 86, "name": "Flashlight", "freebase_id": "/m/01kb5b", "image_count": 60, "instance_count": 85}, {"id": 87, "name": "Billboard", "freebase_id": "/m/01knjb", "image_count": 2628, "instance_count": 6604}, {"id": 88, "name": "Tiara", "freebase_id": "/m/01krhy", "image_count": 330, "instance_count": 360}, {"id": 89, "name": "Limousine", "freebase_id": "/m/01lcw4", "image_count": 315, "instance_count": 346}, {"id": 90, "name": "Necklace", "freebase_id": "/m/01llwg", "image_count": 2122, "instance_count": 2537}, {"id": 91, "name": "Carnivore", "freebase_id": "/m/01lrl", "image_count": 1909, "instance_count": 2859}, {"id": 92, "name": "Scissors", "freebase_id": "/m/01lsmm", "image_count": 283, "instance_count": 388}, {"id": 93, "name": "Stairs", "freebase_id": "/m/01lynh", "image_count": 2057, "instance_count": 2501}, {"id": 94, "name": "Computer keyboard", "freebase_id": "/m/01m2v", "image_count": 3226, "instance_count": 4338}, {"id": 95, "name": "Printer", "freebase_id": "/m/01m4t", "image_count": 199, "instance_count": 252}, {"id": 96, "name": "Traffic sign", "freebase_id": "/m/01mqdt", "image_count": 1923, "instance_count": 4030}, {"id": 97, "name": "Chair", "freebase_id": "/m/01mzpv", "image_count": 19112, "instance_count": 87999}, {"id": 98, "name": "Shirt", "freebase_id": "/m/01n4qj", "image_count": 4971, "instance_count": 7016}, {"id": 99, "name": "Poster", "freebase_id": "/m/01n5jq", "image_count": 14231, "instance_count": 19811}, {"id": 100, "name": "Cheese", "freebase_id": "/m/01nkt", "image_count": 330, "instance_count": 1039}, {"id": 101, "name": "Sock", "freebase_id": "/m/01nq26", "image_count": 643, "instance_count": 1330}, {"id": 102, "name": "Fire hydrant", "freebase_id": "/m/01pns0", "image_count": 388, "instance_count": 412}, {"id": 103, "name": "Land vehicle", "freebase_id": "/m/01prls", "image_count": 20402, "instance_count": 45498}, {"id": 104, "name": "Earrings", "freebase_id": "/m/01r546", "image_count": 661, "instance_count": 1426}, {"id": 105, "name": "Tie", "freebase_id": "/m/01rkbr", "image_count": 7470, "instance_count": 10074}, {"id": 106, "name": "Watercraft", "freebase_id": "/m/01rzcn", "image_count": 2164, "instance_count": 4456}, {"id": 107, "name": "Cabinetry", "freebase_id": "/m/01s105", "image_count": 2011, "instance_count": 5908}, {"id": 108, "name": "Suitcase", "freebase_id": "/m/01s55n", "image_count": 367, "instance_count": 573}, {"id": 109, "name": "Muffin", "freebase_id": "/m/01tcjp", "image_count": 795, "instance_count": 2926}, {"id": 110, "name": "Bidet", "freebase_id": "/m/01vbnl", "image_count": 346, "instance_count": 406}, {"id": 111, "name": "Snack", "freebase_id": "/m/01ww8y", "image_count": 5792, "instance_count": 15326}, {"id": 112, "name": "Snowmobile", "freebase_id": "/m/01x3jk", "image_count": 161, "instance_count": 290}, {"id": 113, "name": "Clock", "freebase_id": "/m/01x3z", "image_count": 846, "instance_count": 1099}, {"id": 114, "name": "Medical equipment", "freebase_id": "/m/01xgg_", "image_count": 588, "instance_count": 1747}, {"id": 115, "name": "Cattle", "freebase_id": "/m/01xq0k1", "image_count": 1786, "instance_count": 6183}, {"id": 116, "name": "Cello", "freebase_id": "/m/01xqw", "image_count": 1230, "instance_count": 1802}, {"id": 117, "name": "Jet ski", "freebase_id": "/m/01xs3r", "image_count": 445, "instance_count": 518}, {"id": 118, "name": "Camel", "freebase_id": "/m/01x_v", "image_count": 632, "instance_count": 1136}, {"id": 119, "name": "Coat", "freebase_id": "/m/01xygc", "image_count": 3005, "instance_count": 5370}, {"id": 120, "name": "Suit", "freebase_id": "/m/01xyhv", "image_count": 39016, "instance_count": 91879}, {"id": 121, "name": "Desk", "freebase_id": "/m/01y9k5", "image_count": 8004, "instance_count": 9623}, {"id": 122, "name": "Cat", "freebase_id": "/m/01yrx", "image_count": 11569, "instance_count": 13979}, {"id": 123, "name": "Bronze sculpture", "freebase_id": "/m/01yx86", "image_count": 1928, "instance_count": 2312}, {"id": 124, "name": "Juice", "freebase_id": "/m/01z1kdw", "image_count": 1500, "instance_count": 2262}, {"id": 125, "name": "Gondola", "freebase_id": "/m/02068x", "image_count": 430, "instance_count": 1375}, {"id": 126, "name": "Beetle", "freebase_id": "/m/020jm", "image_count": 1763, "instance_count": 3256}, {"id": 127, "name": "Cannon", "freebase_id": "/m/020kz", "image_count": 612, "instance_count": 840}, {"id": 128, "name": "Computer mouse", "freebase_id": "/m/020lf", "image_count": 614, "instance_count": 724}, {"id": 129, "name": "Cookie", "freebase_id": "/m/021mn", "image_count": 552, "instance_count": 3012}, {"id": 130, "name": "Office building", "freebase_id": "/m/021sj1", "image_count": 4298, "instance_count": 4331}, {"id": 131, "name": "Fountain", "freebase_id": "/m/0220r2", "image_count": 1888, "instance_count": 2201}, {"id": 132, "name": "Coin", "freebase_id": "/m/0242l", "image_count": 1031, "instance_count": 2533}, {"id": 133, "name": "Calculator", "freebase_id": "/m/024d2", "image_count": 166, "instance_count": 202}, {"id": 134, "name": "Cocktail", "freebase_id": "/m/024g6", "image_count": 2320, "instance_count": 3886}, {"id": 135, "name": "Computer monitor", "freebase_id": "/m/02522", "image_count": 3117, "instance_count": 5551}, {"id": 136, "name": "Box", "freebase_id": "/m/025dyy", "image_count": 1897, "instance_count": 3928}, {"id": 137, "name": "Stapler", "freebase_id": "/m/025fsf", "image_count": 48, "instance_count": 58}, {"id": 138, "name": "Christmas tree", "freebase_id": "/m/025nd", "image_count": 2860, "instance_count": 3816}, {"id": 139, "name": "Cowboy hat", "freebase_id": "/m/025rp__", "image_count": 1727, "instance_count": 2430}, {"id": 140, "name": "Hiking equipment", "freebase_id": "/m/0268lbt", "image_count": 1522, "instance_count": 5428}, {"id": 141, "name": "Studio couch", "freebase_id": "/m/026qbn5", "image_count": 1010, "instance_count": 1694}, {"id": 142, "name": "Drum", "freebase_id": "/m/026t6", "image_count": 6693, "instance_count": 20291}, {"id": 143, "name": "Dessert", "freebase_id": "/m/0270h", "image_count": 6296, "instance_count": 15225}, {"id": 144, "name": "Wine rack", "freebase_id": "/m/0271qf7", "image_count": 92, "instance_count": 178}, {"id": 145, "name": "Drink", "freebase_id": "/m/0271t", "image_count": 13243, "instance_count": 29099}, {"id": 146, "name": "Zucchini", "freebase_id": "/m/027pcv", "image_count": 135, "instance_count": 442}, {"id": 147, "name": "Ladle", "freebase_id": "/m/027rl48", "image_count": 19, "instance_count": 53}, {"id": 148, "name": "Human mouth", "freebase_id": "/m/0283dt1", "image_count": 20878, "instance_count": 33204}, {"id": 149, "name": "Dairy Product", "freebase_id": "/m/0284d", "image_count": 2532, "instance_count": 5946}, {"id": 150, "name": "Dice", "freebase_id": "/m/029b3", "image_count": 180, "instance_count": 557}, {"id": 151, "name": "Oven", "freebase_id": "/m/029bxz", "image_count": 383, "instance_count": 477}, {"id": 152, "name": "Dinosaur", "freebase_id": "/m/029tx", "image_count": 947, "instance_count": 1464}, {"id": 153, "name": "Ratchet (Device)", "freebase_id": "/m/02bm9n", "image_count": 35, "instance_count": 200}, {"id": 154, "name": "Couch", "freebase_id": "/m/02crq1", "image_count": 3017, "instance_count": 3847}, {"id": 155, "name": "Cricket ball", "freebase_id": "/m/02ctlc", "image_count": 83, "instance_count": 107}, {"id": 156, "name": "Winter melon", "freebase_id": "/m/02cvgx", "image_count": 31, "instance_count": 31}, {"id": 157, "name": "Spatula", "freebase_id": "/m/02d1br", "image_count": 33, "instance_count": 50}, {"id": 158, "name": "Whiteboard", "freebase_id": "/m/02d9qx", "image_count": 885, "instance_count": 983}, {"id": 159, "name": "Pencil sharpener", "freebase_id": "/m/02ddwp", "image_count": 13, "instance_count": 21}, {"id": 160, "name": "Door", "freebase_id": "/m/02dgv", "image_count": 11793, "instance_count": 16893}, {"id": 161, "name": "Hat", "freebase_id": "/m/02dl1y", "image_count": 7366, "instance_count": 10747}, {"id": 162, "name": "Shower", "freebase_id": "/m/02f9f_", "image_count": 200, "instance_count": 229}, {"id": 163, "name": "Eraser", "freebase_id": "/m/02fh7f", "image_count": 24, "instance_count": 53}, {"id": 164, "name": "Fedora", "freebase_id": "/m/02fq_6", "image_count": 2660, "instance_count": 3231}, {"id": 165, "name": "Guacamole", "freebase_id": "/m/02g30s", "image_count": 156, "instance_count": 195}, {"id": 166, "name": "Dagger", "freebase_id": "/m/02gzp", "image_count": 258, "instance_count": 363}, {"id": 167, "name": "Scarf", "freebase_id": "/m/02h19r", "image_count": 1843, "instance_count": 2097}, {"id": 168, "name": "Dolphin", "freebase_id": "/m/02hj4", "image_count": 796, "instance_count": 1363}, {"id": 169, "name": "Sombrero", "freebase_id": "/m/02jfl0", "image_count": 439, "instance_count": 561}, {"id": 170, "name": "Tin can", "freebase_id": "/m/02jnhm", "image_count": 1059, "instance_count": 2175}, {"id": 171, "name": "Mug", "freebase_id": "/m/02jvh9", "image_count": 1466, "instance_count": 1969}, {"id": 172, "name": "Tap", "freebase_id": "/m/02jz0l", "image_count": 1031, "instance_count": 1463}, {"id": 173, "name": "Harbor seal", "freebase_id": "/m/02l8p9", "image_count": 745, "instance_count": 1319}, {"id": 174, "name": "Stretcher", "freebase_id": "/m/02lbcq", "image_count": 112, "instance_count": 163}, {"id": 175, "name": "Can opener", "freebase_id": "/m/02mqfb", "image_count": 6, "instance_count": 7}, {"id": 176, "name": "Goggles", "freebase_id": "/m/02_n6y", "image_count": 6447, "instance_count": 8292}, {"id": 177, "name": "Human body", "freebase_id": "/m/02p0tk3", "image_count": 34502, "instance_count": 102243}, {"id": 178, "name": "Roller skates", "freebase_id": "/m/02p3w7d", "image_count": 477, "instance_count": 3430}, {"id": 179, "name": "Coffee cup", "freebase_id": "/m/02p5f1q", "image_count": 3391, "instance_count": 4561}, {"id": 180, "name": "Cutting board", "freebase_id": "/m/02pdsw", "image_count": 167, "instance_count": 180}, {"id": 181, "name": "Blender", "freebase_id": "/m/02pjr4", "image_count": 133, "instance_count": 208}, {"id": 182, "name": "Plumbing fixture", "freebase_id": "/m/02pkr5", "image_count": 251, "instance_count": 461}, {"id": 183, "name": "Stop sign", "freebase_id": "/m/02pv19", "image_count": 318, "instance_count": 330}, {"id": 184, "name": "Office supplies", "freebase_id": "/m/02rdsp", "image_count": 1185, "instance_count": 5501}, {"id": 185, "name": "Volleyball (Ball)", "freebase_id": "/m/02rgn06", "image_count": 539, "instance_count": 610}, {"id": 186, "name": "Vase", "freebase_id": "/m/02s195", "image_count": 1067, "instance_count": 1653}, {"id": 187, "name": "Slow cooker", "freebase_id": "/m/02tsc9", "image_count": 89, "instance_count": 102}, {"id": 188, "name": "Wardrobe", "freebase_id": "/m/02vkqh8", "image_count": 169, "instance_count": 194}, {"id": 189, "name": "Coffee", "freebase_id": "/m/02vqfm", "image_count": 1731, "instance_count": 2060}, {"id": 190, "name": "Whisk", "freebase_id": "/m/02vwcm", "image_count": 157, "instance_count": 180}, {"id": 191, "name": "Paper towel", "freebase_id": "/m/02w3r3", "image_count": 123, "instance_count": 175}, {"id": 192, "name": "Personal care", "freebase_id": "/m/02w3_ws", "image_count": 115, "instance_count": 358}, {"id": 193, "name": "Food", "freebase_id": "/m/02wbm", "image_count": 15714, "instance_count": 43473}, {"id": 194, "name": "Sun hat", "freebase_id": "/m/02wbtzl", "image_count": 3547, "instance_count": 5325}, {"id": 195, "name": "Tree house", "freebase_id": "/m/02wg_p", "image_count": 85, "instance_count": 87}, {"id": 196, "name": "Flying disc", "freebase_id": "/m/02wmf", "image_count": 178, "instance_count": 201}, {"id": 197, "name": "Skirt", "freebase_id": "/m/02wv6h6", "image_count": 819, "instance_count": 1060}, {"id": 198, "name": "Gas stove", "freebase_id": "/m/02wv84t", "image_count": 369, "instance_count": 412}, {"id": 199, "name": "Salt and pepper shakers", "freebase_id": "/m/02x8cch", "image_count": 89, "instance_count": 167}, {"id": 200, "name": "Mechanical fan", "freebase_id": "/m/02x984l", "image_count": 542, "instance_count": 626}, {"id": 201, "name": "Face powder", "freebase_id": "/m/02xb7qb", "image_count": 45, "instance_count": 72}, {"id": 202, "name": "Fax", "freebase_id": "/m/02xqq", "image_count": 22, "instance_count": 28}, {"id": 203, "name": "Fruit", "freebase_id": "/m/02xwb", "image_count": 2074, "instance_count": 6691}, {"id": 204, "name": "French fries", "freebase_id": "/m/02y6n", "image_count": 64, "instance_count": 598}, {"id": 205, "name": "Nightstand", "freebase_id": "/m/02z51p", "image_count": 620, "instance_count": 917}, {"id": 206, "name": "Barrel", "freebase_id": "/m/02zn6n", "image_count": 368, "instance_count": 1227}, {"id": 207, "name": "Kite", "freebase_id": "/m/02zt3", "image_count": 313, "instance_count": 1373}, {"id": 208, "name": "Tart", "freebase_id": "/m/02zvsm", "image_count": 315, "instance_count": 579}, {"id": 209, "name": "Treadmill", "freebase_id": "/m/030610", "image_count": 90, "instance_count": 190}, {"id": 210, "name": "Fox", "freebase_id": "/m/0306r", "image_count": 404, "instance_count": 462}, {"id": 211, "name": "Flag", "freebase_id": "/m/03120", "image_count": 6289, "instance_count": 18394}, {"id": 212, "name": "French horn", "freebase_id": "/m/0319l", "image_count": 264, "instance_count": 874}, {"id": 213, "name": "Window blind", "freebase_id": "/m/031b6r", "image_count": 378, "instance_count": 535}, {"id": 214, "name": "Human foot", "freebase_id": "/m/031n1", "image_count": 920, "instance_count": 2034}, {"id": 215, "name": "Golf cart", "freebase_id": "/m/0323sq", "image_count": 240, "instance_count": 375}, {"id": 216, "name": "Jacket", "freebase_id": "/m/032b3c", "image_count": 10693, "instance_count": 20010}, {"id": 217, "name": "Egg (Food)", "freebase_id": "/m/033cnk", "image_count": 514, "instance_count": 1026}, {"id": 218, "name": "Street light", "freebase_id": "/m/033rq4", "image_count": 7025, "instance_count": 24201}, {"id": 219, "name": "Guitar", "freebase_id": "/m/0342h", "image_count": 16260, "instance_count": 23053}, {"id": 220, "name": "Pillow", "freebase_id": "/m/034c16", "image_count": 1183, "instance_count": 3280}, {"id": 221, "name": "Human leg", "freebase_id": "/m/035r7c", "image_count": 13140, "instance_count": 47109}, {"id": 222, "name": "Isopod", "freebase_id": "/m/035vxb", "image_count": 85, "instance_count": 141}, {"id": 223, "name": "Grape", "freebase_id": "/m/0388q", "image_count": 20, "instance_count": 161}, {"id": 224, "name": "Human ear", "freebase_id": "/m/039xj_", "image_count": 7339, "instance_count": 13645}, {"id": 225, "name": "Power plugs and sockets", "freebase_id": "/m/03bbps", "image_count": 139, "instance_count": 272}, {"id": 226, "name": "Panda", "freebase_id": "/m/03bj1", "image_count": 480, "instance_count": 706}, {"id": 227, "name": "Giraffe", "freebase_id": "/m/03bk1", "image_count": 680, "instance_count": 1024}, {"id": 228, "name": "Woman", "freebase_id": "/m/03bt1vf", "image_count": 211054, "instance_count": 519938}, {"id": 229, "name": "Door handle", "freebase_id": "/m/03c7gz", "image_count": 557, "instance_count": 716}, {"id": 230, "name": "Rhinoceros", "freebase_id": "/m/03d443", "image_count": 447, "instance_count": 611}, {"id": 231, "name": "Bathtub", "freebase_id": "/m/03dnzn", "image_count": 490, "instance_count": 501}, {"id": 232, "name": "Goldfish", "freebase_id": "/m/03fj2", "image_count": 325, "instance_count": 1409}, {"id": 233, "name": "Houseplant", "freebase_id": "/m/03fp41", "image_count": 5512, "instance_count": 11884}, {"id": 234, "name": "Goat", "freebase_id": "/m/03fwl", "image_count": 827, "instance_count": 1418}, {"id": 235, "name": "Baseball bat", "freebase_id": "/m/03g8mr", "image_count": 919, "instance_count": 1031}, {"id": 236, "name": "Baseball glove", "freebase_id": "/m/03grzl", "image_count": 1423, "instance_count": 2114}, {"id": 237, "name": "Mixing bowl", "freebase_id": "/m/03hj559", "image_count": 480, "instance_count": 761}, {"id": 238, "name": "Marine invertebrates", "freebase_id": "/m/03hl4l9", "image_count": 1335, "instance_count": 7073}, {"id": 239, "name": "Kitchen utensil", "freebase_id": "/m/03hlz0c", "image_count": 208, "instance_count": 390}, {"id": 240, "name": "Light switch", "freebase_id": "/m/03jbxj", "image_count": 78, "instance_count": 95}, {"id": 241, "name": "House", "freebase_id": "/m/03jm5", "image_count": 37094, "instance_count": 81831}, {"id": 242, "name": "Horse", "freebase_id": "/m/03k3r", "image_count": 5023, "instance_count": 9118}, {"id": 243, "name": "Stationary bicycle", "freebase_id": "/m/03kt2w", "image_count": 81, "instance_count": 226}, {"id": 244, "name": "Hammer", "freebase_id": "/m/03l9g", "image_count": 117, "instance_count": 136}, {"id": 245, "name": "Ceiling fan", "freebase_id": "/m/03ldnb", "image_count": 399, "instance_count": 442}, {"id": 246, "name": "Sofa bed", "freebase_id": "/m/03m3pdh", "image_count": 1031, "instance_count": 1360}, {"id": 247, "name": "Adhesive tape", "freebase_id": "/m/03m3vtv", "image_count": 94, "instance_count": 203}, {"id": 248, "name": "Harp", "freebase_id": "/m/03m5k", "image_count": 197, "instance_count": 227}, {"id": 249, "name": "Sandal", "freebase_id": "/m/03nfch", "image_count": 1214, "instance_count": 2617}, {"id": 250, "name": "Bicycle helmet", "freebase_id": "/m/03p3bw", "image_count": 2970, "instance_count": 6430}, {"id": 251, "name": "Saucer", "freebase_id": "/m/03q5c7", "image_count": 1388, "instance_count": 2349}, {"id": 252, "name": "Harpsichord", "freebase_id": "/m/03q5t", "image_count": 189, "instance_count": 206}, {"id": 253, "name": "Human hair", "freebase_id": "/m/03q69", "image_count": 43629, "instance_count": 125286}, {"id": 254, "name": "Heater", "freebase_id": "/m/03qhv5", "image_count": 32, "instance_count": 35}, {"id": 255, "name": "Harmonica", "freebase_id": "/m/03qjg", "image_count": 20, "instance_count": 38}, {"id": 256, "name": "Hamster", "freebase_id": "/m/03qrc", "image_count": 433, "instance_count": 518}, {"id": 257, "name": "Curtain", "freebase_id": "/m/03rszm", "image_count": 2271, "instance_count": 4094}, {"id": 258, "name": "Bed", "freebase_id": "/m/03ssj5", "image_count": 2667, "instance_count": 3172}, {"id": 259, "name": "Kettle", "freebase_id": "/m/03s_tn", "image_count": 418, "instance_count": 562}, {"id": 260, "name": "Fireplace", "freebase_id": "/m/03tw93", "image_count": 659, "instance_count": 685}, {"id": 261, "name": "Scale", "freebase_id": "/m/03txqz", "image_count": 130, "instance_count": 138}, {"id": 262, "name": "Drinking straw", "freebase_id": "/m/03v5tg", "image_count": 140, "instance_count": 236}, {"id": 263, "name": "Insect", "freebase_id": "/m/03vt0", "image_count": 5504, "instance_count": 7494}, {"id": 264, "name": "Hair dryer", "freebase_id": "/m/03wvsk", "image_count": 27, "instance_count": 27}, {"id": 265, "name": "Kitchenware", "freebase_id": "/m/03_wxk", "image_count": 0, "instance_count": 0}, {"id": 266, "name": "Indoor rower", "freebase_id": "/m/03wym", "image_count": 12, "instance_count": 13}, {"id": 267, "name": "Invertebrate", "freebase_id": "/m/03xxp", "image_count": 684, "instance_count": 1411}, {"id": 268, "name": "Food processor", "freebase_id": "/m/03y6mg", "image_count": 119, "instance_count": 176}, {"id": 269, "name": "Bookcase", "freebase_id": "/m/03__z0", "image_count": 781, "instance_count": 1155}, {"id": 270, "name": "Refrigerator", "freebase_id": "/m/040b_t", "image_count": 467, "instance_count": 506}, {"id": 271, "name": "Wood-burning stove", "freebase_id": "/m/04169hn", "image_count": 276, "instance_count": 289}, {"id": 272, "name": "Punching bag", "freebase_id": "/m/0420v5", "image_count": 117, "instance_count": 329}, {"id": 273, "name": "Common fig", "freebase_id": "/m/043nyj", "image_count": 65, "instance_count": 212}, {"id": 274, "name": "Cocktail shaker", "freebase_id": "/m/0440zs", "image_count": 23, "instance_count": 24}, {"id": 275, "name": "Jaguar (Animal)", "freebase_id": "/m/0449p", "image_count": 410, "instance_count": 468}, {"id": 276, "name": "Golf ball", "freebase_id": "/m/044r5d", "image_count": 279, "instance_count": 338}, {"id": 277, "name": "Fashion accessory", "freebase_id": "/m/0463sg", "image_count": 19718, "instance_count": 72153}, {"id": 278, "name": "Alarm clock", "freebase_id": "/m/046dlr", "image_count": 129, "instance_count": 162}, {"id": 279, "name": "Filing cabinet", "freebase_id": "/m/047j0r", "image_count": 119, "instance_count": 143}, {"id": 280, "name": "Artichoke", "freebase_id": "/m/047v4b", "image_count": 107, "instance_count": 217}, {"id": 281, "name": "Table", "freebase_id": "/m/04bcr3", "image_count": 35125, "instance_count": 63501}, {"id": 282, "name": "Tableware", "freebase_id": "/m/04brg2", "image_count": 3477, "instance_count": 20098}, {"id": 283, "name": "Kangaroo", "freebase_id": "/m/04c0y", "image_count": 379, "instance_count": 625}, {"id": 284, "name": "Koala", "freebase_id": "/m/04cp_", "image_count": 337, "instance_count": 379}, {"id": 285, "name": "Knife", "freebase_id": "/m/04ctx", "image_count": 569, "instance_count": 781}, {"id": 286, "name": "Bottle", "freebase_id": "/m/04dr76w", "image_count": 9697, "instance_count": 24636}, {"id": 287, "name": "Bottle opener", "freebase_id": "/m/04f5ws", "image_count": 17, "instance_count": 21}, {"id": 288, "name": "Lynx", "freebase_id": "/m/04g2r", "image_count": 191, "instance_count": 204}, {"id": 289, "name": "Lavender (Plant)", "freebase_id": "/m/04gth", "image_count": 230, "instance_count": 1711}, {"id": 290, "name": "Lighthouse", "freebase_id": "/m/04h7h", "image_count": 1285, "instance_count": 1334}, {"id": 291, "name": "Dumbbell", "freebase_id": "/m/04h8sr", "image_count": 82, "instance_count": 254}, {"id": 292, "name": "Human head", "freebase_id": "/m/04hgtk", "image_count": 39752, "instance_count": 109981}, {"id": 293, "name": "Bowl", "freebase_id": "/m/04kkgm", "image_count": 1178, "instance_count": 2754}, {"id": 294, "name": "Humidifier", "freebase_id": "/m/04lvq_", "image_count": 9, "instance_count": 11}, {"id": 295, "name": "Porch", "freebase_id": "/m/04m6gz", "image_count": 1740, "instance_count": 2379}, {"id": 296, "name": "Lizard", "freebase_id": "/m/04m9y", "image_count": 1702, "instance_count": 1954}, {"id": 297, "name": "Billiard table", "freebase_id": "/m/04p0qw", "image_count": 563, "instance_count": 765}, {"id": 298, "name": "Mammal", "freebase_id": "/m/04rky", "image_count": 23173, "instance_count": 81163}, {"id": 299, "name": "Mouse", "freebase_id": "/m/04rmv", "image_count": 662, "instance_count": 797}, {"id": 300, "name": "Motorcycle", "freebase_id": "/m/04_sv", "image_count": 5325, "instance_count": 9382}, {"id": 301, "name": "Musical instrument", "freebase_id": "/m/04szw", "image_count": 4253, "instance_count": 10878}, {"id": 302, "name": "Swim cap", "freebase_id": "/m/04tn4x", "image_count": 303, "instance_count": 584}, {"id": 303, "name": "Frying pan", "freebase_id": "/m/04v6l4", "image_count": 260, "instance_count": 324}, {"id": 304, "name": "Snowplow", "freebase_id": "/m/04vv5k", "image_count": 196, "instance_count": 252}, {"id": 305, "name": "Bathroom cabinet", "freebase_id": "/m/04y4h8h", "image_count": 256, "instance_count": 277}, {"id": 306, "name": "Missile", "freebase_id": "/m/04ylt", "image_count": 420, "instance_count": 581}, {"id": 307, "name": "Bust", "freebase_id": "/m/04yqq2", "image_count": 843, "instance_count": 985}, {"id": 308, "name": "Man", "freebase_id": "/m/04yx4", "image_count": 298992, "instance_count": 958133}, {"id": 309, "name": "Waffle iron", "freebase_id": "/m/04z4wx", "image_count": 22, "instance_count": 29}, {"id": 310, "name": "Milk", "freebase_id": "/m/04zpv", "image_count": 144, "instance_count": 195}, {"id": 311, "name": "Ring binder", "freebase_id": "/m/04zwwv", "image_count": 47, "instance_count": 82}, {"id": 312, "name": "Plate", "freebase_id": "/m/050gv4", "image_count": 1643, "instance_count": 3780}, {"id": 313, "name": "Mobile phone", "freebase_id": "/m/050k8", "image_count": 4145, "instance_count": 6051}, {"id": 314, "name": "Baked goods", "freebase_id": "/m/052lwg6", "image_count": 7485, "instance_count": 14300}, {"id": 315, "name": "Mushroom", "freebase_id": "/m/052sf", "image_count": 814, "instance_count": 1951}, {"id": 316, "name": "Crutch", "freebase_id": "/m/05441v", "image_count": 74, "instance_count": 136}, {"id": 317, "name": "Pitcher (Container)", "freebase_id": "/m/054fyh", "image_count": 252, "instance_count": 303}, {"id": 318, "name": "Mirror", "freebase_id": "/m/054_l", "image_count": 971, "instance_count": 1348}, {"id": 319, "name": "Personal flotation device", "freebase_id": "/m/054xkw", "image_count": 1025, "instance_count": 2793}, {"id": 320, "name": "Table tennis racket", "freebase_id": "/m/05_5p_0", "image_count": 566, "instance_count": 815}, {"id": 321, "name": "Pencil case", "freebase_id": "/m/05676x", "image_count": 89, "instance_count": 116}, {"id": 322, "name": "Musical keyboard", "freebase_id": "/m/057cc", "image_count": 1435, "instance_count": 1682}, {"id": 323, "name": "Scoreboard", "freebase_id": "/m/057p5t", "image_count": 347, "instance_count": 376}, {"id": 324, "name": "Briefcase", "freebase_id": "/m/0584n8", "image_count": 106, "instance_count": 135}, {"id": 325, "name": "Kitchen knife", "freebase_id": "/m/058qzx", "image_count": 240, "instance_count": 300}, {"id": 326, "name": "Nail (Construction)", "freebase_id": "/m/05bm6", "image_count": 118, "instance_count": 480}, {"id": 327, "name": "Tennis ball", "freebase_id": "/m/05ctyq", "image_count": 345, "instance_count": 426}, {"id": 328, "name": "Plastic bag", "freebase_id": "/m/05gqfk", "image_count": 391, "instance_count": 682}, {"id": 329, "name": "Oboe", "freebase_id": "/m/05kms", "image_count": 86, "instance_count": 137}, {"id": 330, "name": "Chest of drawers", "freebase_id": "/m/05kyg_", "image_count": 515, "instance_count": 650}, {"id": 331, "name": "Ostrich", "freebase_id": "/m/05n4y", "image_count": 303, "instance_count": 487}, {"id": 332, "name": "Piano", "freebase_id": "/m/05r5c", "image_count": 1184, "instance_count": 1301}, {"id": 333, "name": "Girl", "freebase_id": "/m/05r655", "image_count": 72741, "instance_count": 135861}, {"id": 334, "name": "Plant", "freebase_id": "/m/05s2s", "image_count": 17891, "instance_count": 62731}, {"id": 335, "name": "Potato", "freebase_id": "/m/05vtc", "image_count": 136, "instance_count": 274}, {"id": 336, "name": "Hair spray", "freebase_id": "/m/05w9t9", "image_count": 5, "instance_count": 9}, {"id": 337, "name": "Sports equipment", "freebase_id": "/m/05y5lj", "image_count": 6388, "instance_count": 25243}, {"id": 338, "name": "Pasta", "freebase_id": "/m/05z55", "image_count": 646, "instance_count": 654}, {"id": 339, "name": "Penguin", "freebase_id": "/m/05z6w", "image_count": 768, "instance_count": 3036}, {"id": 340, "name": "Pumpkin", "freebase_id": "/m/05zsy", "image_count": 1126, "instance_count": 3110}, {"id": 341, "name": "Pear", "freebase_id": "/m/061_f", "image_count": 224, "instance_count": 653}, {"id": 342, "name": "Infant bed", "freebase_id": "/m/061hd_", "image_count": 411, "instance_count": 444}, {"id": 343, "name": "Polar bear", "freebase_id": "/m/0633h", "image_count": 471, "instance_count": 630}, {"id": 344, "name": "Mixer", "freebase_id": "/m/063rgb", "image_count": 173, "instance_count": 207}, {"id": 345, "name": "Cupboard", "freebase_id": "/m/0642b4", "image_count": 531, "instance_count": 744}, {"id": 346, "name": "Jacuzzi", "freebase_id": "/m/065h6l", "image_count": 93, "instance_count": 94}, {"id": 347, "name": "Pizza", "freebase_id": "/m/0663v", "image_count": 1122, "instance_count": 1740}, {"id": 348, "name": "Digital clock", "freebase_id": "/m/06_72j", "image_count": 168, "instance_count": 198}, {"id": 349, "name": "Pig", "freebase_id": "/m/068zj", "image_count": 638, "instance_count": 1169}, {"id": 350, "name": "Reptile", "freebase_id": "/m/06bt6", "image_count": 413, "instance_count": 498}, {"id": 351, "name": "Rifle", "freebase_id": "/m/06c54", "image_count": 1548, "instance_count": 2153}, {"id": 352, "name": "Lipstick", "freebase_id": "/m/06c7f7", "image_count": 1063, "instance_count": 1276}, {"id": 353, "name": "Skateboard", "freebase_id": "/m/06_fw", "image_count": 907, "instance_count": 1485}, {"id": 354, "name": "Raven", "freebase_id": "/m/06j2d", "image_count": 377, "instance_count": 495}, {"id": 355, "name": "High heels", "freebase_id": "/m/06k2mb", "image_count": 1451, "instance_count": 2887}, {"id": 356, "name": "Red panda", "freebase_id": "/m/06l9r", "image_count": 287, "instance_count": 356}, {"id": 357, "name": "Rose", "freebase_id": "/m/06m11", "image_count": 3674, "instance_count": 8391}, {"id": 358, "name": "Rabbit", "freebase_id": "/m/06mf6", "image_count": 972, "instance_count": 1289}, {"id": 359, "name": "Sculpture", "freebase_id": "/m/06msq", "image_count": 14266, "instance_count": 26382}, {"id": 360, "name": "Saxophone", "freebase_id": "/m/06ncr", "image_count": 746, "instance_count": 980}, {"id": 361, "name": "Shotgun", "freebase_id": "/m/06nrc", "image_count": 406, "instance_count": 543}, {"id": 362, "name": "Seafood", "freebase_id": "/m/06nwz", "image_count": 543, "instance_count": 1021}, {"id": 363, "name": "Submarine sandwich", "freebase_id": "/m/06pcq", "image_count": 192, "instance_count": 222}, {"id": 364, "name": "Snowboard", "freebase_id": "/m/06__v", "image_count": 476, "instance_count": 717}, {"id": 365, "name": "Sword", "freebase_id": "/m/06y5r", "image_count": 385, "instance_count": 552}, {"id": 366, "name": "Picture frame", "freebase_id": "/m/06z37_", "image_count": 4917, "instance_count": 10425}, {"id": 367, "name": "Sushi", "freebase_id": "/m/07030", "image_count": 258, "instance_count": 773}, {"id": 368, "name": "Loveseat", "freebase_id": "/m/0703r8", "image_count": 465, "instance_count": 577}, {"id": 369, "name": "Ski", "freebase_id": "/m/071p9", "image_count": 931, "instance_count": 2075}, {"id": 370, "name": "Squirrel", "freebase_id": "/m/071qp", "image_count": 1506, "instance_count": 1590}, {"id": 371, "name": "Tripod", "freebase_id": "/m/073bxn", "image_count": 1067, "instance_count": 1281}, {"id": 372, "name": "Stethoscope", "freebase_id": "/m/073g6", "image_count": 74, "instance_count": 77}, {"id": 373, "name": "Submarine", "freebase_id": "/m/074d1", "image_count": 74, "instance_count": 80}, {"id": 374, "name": "Scorpion", "freebase_id": "/m/0755b", "image_count": 146, "instance_count": 179}, {"id": 375, "name": "Segway", "freebase_id": "/m/076bq", "image_count": 246, "instance_count": 494}, {"id": 376, "name": "Training bench", "freebase_id": "/m/076lb9", "image_count": 81, "instance_count": 149}, {"id": 377, "name": "Snake", "freebase_id": "/m/078jl", "image_count": 1112, "instance_count": 1292}, {"id": 378, "name": "Coffee table", "freebase_id": "/m/078n6m", "image_count": 2363, "instance_count": 4217}, {"id": 379, "name": "Skyscraper", "freebase_id": "/m/079cl", "image_count": 13904, "instance_count": 48273}, {"id": 380, "name": "Sheep", "freebase_id": "/m/07bgp", "image_count": 677, "instance_count": 1950}, {"id": 381, "name": "Television", "freebase_id": "/m/07c52", "image_count": 2487, "instance_count": 3340}, {"id": 382, "name": "Trombone", "freebase_id": "/m/07c6l", "image_count": 364, "instance_count": 639}, {"id": 383, "name": "Tea", "freebase_id": "/m/07clx", "image_count": 954, "instance_count": 1181}, {"id": 384, "name": "Tank", "freebase_id": "/m/07cmd", "image_count": 873, "instance_count": 1362}, {"id": 385, "name": "Taco", "freebase_id": "/m/07crc", "image_count": 306, "instance_count": 601}, {"id": 386, "name": "Telephone", "freebase_id": "/m/07cx4", "image_count": 211, "instance_count": 261}, {"id": 387, "name": "Torch", "freebase_id": "/m/07dd4", "image_count": 18, "instance_count": 20}, {"id": 388, "name": "Tiger", "freebase_id": "/m/07dm6", "image_count": 797, "instance_count": 963}, {"id": 389, "name": "Strawberry", "freebase_id": "/m/07fbm7", "image_count": 867, "instance_count": 5729}, {"id": 390, "name": "Trumpet", "freebase_id": "/m/07gql", "image_count": 696, "instance_count": 1160}, {"id": 391, "name": "Tree", "freebase_id": "/m/07j7r", "image_count": 106636, "instance_count": 375594}, {"id": 392, "name": "Tomato", "freebase_id": "/m/07j87", "image_count": 762, "instance_count": 3783}, {"id": 393, "name": "Train", "freebase_id": "/m/07jdr", "image_count": 7263, "instance_count": 9467}, {"id": 394, "name": "Tool", "freebase_id": "/m/07k1x", "image_count": 709, "instance_count": 1617}, {"id": 395, "name": "Picnic basket", "freebase_id": "/m/07kng9", "image_count": 241, "instance_count": 348}, {"id": 396, "name": "Cooking spray", "freebase_id": "/m/07mcwg", "image_count": 9, "instance_count": 42}, {"id": 397, "name": "Trousers", "freebase_id": "/m/07mhn", "image_count": 4002, "instance_count": 6654}, {"id": 398, "name": "Bowling equipment", "freebase_id": "/m/07pj7bq", "image_count": 245, "instance_count": 743}, {"id": 399, "name": "Football helmet", "freebase_id": "/m/07qxg_", "image_count": 1269, "instance_count": 5564}, {"id": 400, "name": "Truck", "freebase_id": "/m/07r04", "image_count": 6185, "instance_count": 9046}, {"id": 401, "name": "Measuring cup", "freebase_id": "/m/07v9_z", "image_count": 64, "instance_count": 73}, {"id": 402, "name": "Coffeemaker", "freebase_id": "/m/07xyvk", "image_count": 223, "instance_count": 311}, {"id": 403, "name": "Violin", "freebase_id": "/m/07y_7", "image_count": 1195, "instance_count": 2233}, {"id": 404, "name": "Vehicle", "freebase_id": "/m/07yv9", "image_count": 10989, "instance_count": 27848}, {"id": 405, "name": "Handbag", "freebase_id": "/m/080hkjn", "image_count": 1740, "instance_count": 2248}, {"id": 406, "name": "Paper cutter", "freebase_id": "/m/080n7g", "image_count": 3, "instance_count": 4}, {"id": 407, "name": "Wine", "freebase_id": "/m/081qc", "image_count": 3951, "instance_count": 9670}, {"id": 408, "name": "Weapon", "freebase_id": "/m/083kb", "image_count": 1383, "instance_count": 2360}, {"id": 409, "name": "Wheel", "freebase_id": "/m/083wq", "image_count": 58010, "instance_count": 226764}, {"id": 410, "name": "Worm", "freebase_id": "/m/084hf", "image_count": 183, "instance_count": 242}, {"id": 411, "name": "Wok", "freebase_id": "/m/084rd", "image_count": 330, "instance_count": 511}, {"id": 412, "name": "Whale", "freebase_id": "/m/084zz", "image_count": 672, "instance_count": 909}, {"id": 413, "name": "Zebra", "freebase_id": "/m/0898b", "image_count": 351, "instance_count": 555}, {"id": 414, "name": "Auto part", "freebase_id": "/m/08dz3q", "image_count": 2539, "instance_count": 5156}, {"id": 415, "name": "Jug", "freebase_id": "/m/08hvt4", "image_count": 343, "instance_count": 500}, {"id": 416, "name": "Pizza cutter", "freebase_id": "/m/08ks85", "image_count": 19, "instance_count": 20}, {"id": 417, "name": "Cream", "freebase_id": "/m/08p92x", "image_count": 68, "instance_count": 120}, {"id": 418, "name": "Monkey", "freebase_id": "/m/08pbxl", "image_count": 1774, "instance_count": 2484}, {"id": 419, "name": "Lion", "freebase_id": "/m/096mb", "image_count": 954, "instance_count": 1276}, {"id": 420, "name": "Bread", "freebase_id": "/m/09728", "image_count": 1280, "instance_count": 2639}, {"id": 421, "name": "Platter", "freebase_id": "/m/099ssp", "image_count": 986, "instance_count": 2243}, {"id": 422, "name": "Chicken", "freebase_id": "/m/09b5t", "image_count": 1153, "instance_count": 2229}, {"id": 423, "name": "Eagle", "freebase_id": "/m/09csl", "image_count": 1391, "instance_count": 1591}, {"id": 424, "name": "Helicopter", "freebase_id": "/m/09ct_", "image_count": 2179, "instance_count": 2715}, {"id": 425, "name": "Owl", "freebase_id": "/m/09d5_", "image_count": 1212, "instance_count": 1493}, {"id": 426, "name": "Duck", "freebase_id": "/m/09ddx", "image_count": 3773, "instance_count": 10447}, {"id": 427, "name": "Turtle", "freebase_id": "/m/09dzg", "image_count": 119, "instance_count": 165}, {"id": 428, "name": "Hippopotamus", "freebase_id": "/m/09f20", "image_count": 343, "instance_count": 571}, {"id": 429, "name": "Crocodile", "freebase_id": "/m/09f_2", "image_count": 606, "instance_count": 853}, {"id": 430, "name": "Toilet", "freebase_id": "/m/09g1w", "image_count": 811, "instance_count": 1033}, {"id": 431, "name": "Toilet paper", "freebase_id": "/m/09gtd", "image_count": 195, "instance_count": 313}, {"id": 432, "name": "Squid", "freebase_id": "/m/09gys", "image_count": 78, "instance_count": 137}, {"id": 433, "name": "Clothing", "freebase_id": "/m/09j2d", "image_count": 253553, "instance_count": 937287}, {"id": 434, "name": "Footwear", "freebase_id": "/m/09j5n", "image_count": 94846, "instance_count": 484352}, {"id": 435, "name": "Lemon", "freebase_id": "/m/09k_b", "image_count": 490, "instance_count": 1182}, {"id": 436, "name": "Spider", "freebase_id": "/m/09kmb", "image_count": 1621, "instance_count": 1930}, {"id": 437, "name": "Deer", "freebase_id": "/m/09kx5", "image_count": 1353, "instance_count": 2310}, {"id": 438, "name": "Frog", "freebase_id": "/m/09ld4", "image_count": 1248, "instance_count": 1465}, {"id": 439, "name": "Banana", "freebase_id": "/m/09qck", "image_count": 358, "instance_count": 667}, {"id": 440, "name": "Rocket", "freebase_id": "/m/09rvcxw", "image_count": 622, "instance_count": 890}, {"id": 441, "name": "Wine glass", "freebase_id": "/m/09tvcd", "image_count": 4101, "instance_count": 10326}, {"id": 442, "name": "Countertop", "freebase_id": "/m/0b3fp9", "image_count": 1608, "instance_count": 2240}, {"id": 443, "name": "Tablet computer", "freebase_id": "/m/0bh9flk", "image_count": 756, "instance_count": 936}, {"id": 444, "name": "Waste container", "freebase_id": "/m/0bjyj5", "image_count": 853, "instance_count": 1474}, {"id": 445, "name": "Swimming pool", "freebase_id": "/m/0b_rs", "image_count": 2872, "instance_count": 3057}, {"id": 446, "name": "Dog", "freebase_id": "/m/0bt9lr", "image_count": 18057, "instance_count": 24721}, {"id": 447, "name": "Book", "freebase_id": "/m/0bt_c3", "image_count": 3640, "instance_count": 7698}, {"id": 448, "name": "Elephant", "freebase_id": "/m/0bwd_0j", "image_count": 1225, "instance_count": 2263}, {"id": 449, "name": "Shark", "freebase_id": "/m/0by6g", "image_count": 486, "instance_count": 613}, {"id": 450, "name": "Candle", "freebase_id": "/m/0c06p", "image_count": 1005, "instance_count": 4531}, {"id": 451, "name": "Leopard", "freebase_id": "/m/0c29q", "image_count": 534, "instance_count": 613}, {"id": 452, "name": "Axe", "freebase_id": "/m/0c2jj", "image_count": 112, "instance_count": 145}, {"id": 453, "name": "Hand dryer", "freebase_id": "/m/0c3m8g", "image_count": 56, "instance_count": 70}, {"id": 454, "name": "Soap dispenser", "freebase_id": "/m/0c3mkw", "image_count": 58, "instance_count": 78}, {"id": 455, "name": "Porcupine", "freebase_id": "/m/0c568", "image_count": 182, "instance_count": 205}, {"id": 456, "name": "Flower", "freebase_id": "/m/0c9ph5", "image_count": 29498, "instance_count": 125304}, {"id": 457, "name": "Canary", "freebase_id": "/m/0ccs93", "image_count": 206, "instance_count": 322}, {"id": 458, "name": "Cheetah", "freebase_id": "/m/0cd4d", "image_count": 424, "instance_count": 533}, {"id": 459, "name": "Palm tree", "freebase_id": "/m/0cdl1", "image_count": 7987, "instance_count": 31443}, {"id": 460, "name": "Hamburger", "freebase_id": "/m/0cdn1", "image_count": 624, "instance_count": 835}, {"id": 461, "name": "Maple", "freebase_id": "/m/0cffdh", "image_count": 1628, "instance_count": 3515}, {"id": 462, "name": "Building", "freebase_id": "/m/0cgh4", "image_count": 64125, "instance_count": 102491}, {"id": 463, "name": "Fish", "freebase_id": "/m/0ch_cf", "image_count": 4327, "instance_count": 14862}, {"id": 464, "name": "Lobster", "freebase_id": "/m/0cjq5", "image_count": 261, "instance_count": 384}, {"id": 465, "name": "Garden Asparagus", "freebase_id": "/m/0cjs7", "image_count": 59, "instance_count": 84}, {"id": 466, "name": "Furniture", "freebase_id": "/m/0c_jw", "image_count": 11135, "instance_count": 27884}, {"id": 467, "name": "Hedgehog", "freebase_id": "/m/0cl4p", "image_count": 213, "instance_count": 235}, {"id": 468, "name": "Airplane", "freebase_id": "/m/0cmf2", "image_count": 11295, "instance_count": 19823}, {"id": 469, "name": "Spoon", "freebase_id": "/m/0cmx8", "image_count": 913, "instance_count": 1435}, {"id": 470, "name": "Otter", "freebase_id": "/m/0cn6p", "image_count": 439, "instance_count": 661}, {"id": 471, "name": "Bull", "freebase_id": "/m/0cnyhnx", "image_count": 829, "instance_count": 1187}, {"id": 472, "name": "Oyster", "freebase_id": "/m/0_cp5", "image_count": 235, "instance_count": 603}, {"id": 473, "name": "Horizontal bar", "freebase_id": "/m/0cqn2", "image_count": 48, "instance_count": 64}, {"id": 474, "name": "Convenience store", "freebase_id": "/m/0crjs", "image_count": 1233, "instance_count": 1240}, {"id": 475, "name": "Bomb", "freebase_id": "/m/0ct4f", "image_count": 6, "instance_count": 8}, {"id": 476, "name": "Bench", "freebase_id": "/m/0cvnqh", "image_count": 2085, "instance_count": 3801}, {"id": 477, "name": "Ice cream", "freebase_id": "/m/0cxn2", "image_count": 1331, "instance_count": 2451}, {"id": 478, "name": "Caterpillar", "freebase_id": "/m/0cydv", "image_count": 622, "instance_count": 779}, {"id": 479, "name": "Butterfly", "freebase_id": "/m/0cyf8", "image_count": 3871, "instance_count": 8739}, {"id": 480, "name": "Parachute", "freebase_id": "/m/0cyfs", "image_count": 974, "instance_count": 2342}, {"id": 481, "name": "Orange", "freebase_id": "/m/0cyhj_", "image_count": 653, "instance_count": 3820}, {"id": 482, "name": "Antelope", "freebase_id": "/m/0czz2", "image_count": 602, "instance_count": 1074}, {"id": 483, "name": "Beaker", "freebase_id": "/m/0d20w4", "image_count": 84, "instance_count": 165}, {"id": 484, "name": "Moths and butterflies", "freebase_id": "/m/0d_2m", "image_count": 1650, "instance_count": 1805}, {"id": 485, "name": "Window", "freebase_id": "/m/0d4v4", "image_count": 42733, "instance_count": 306327}, {"id": 486, "name": "Closet", "freebase_id": "/m/0d4w1", "image_count": 259, "instance_count": 452}, {"id": 487, "name": "Castle", "freebase_id": "/m/0d5gx", "image_count": 2649, "instance_count": 2987}, {"id": 488, "name": "Jellyfish", "freebase_id": "/m/0d8zb", "image_count": 632, "instance_count": 1626}, {"id": 489, "name": "Goose", "freebase_id": "/m/0dbvp", "image_count": 2052, "instance_count": 5910}, {"id": 490, "name": "Mule", "freebase_id": "/m/0dbzx", "image_count": 427, "instance_count": 770}, {"id": 491, "name": "Swan", "freebase_id": "/m/0dftk", "image_count": 1105, "instance_count": 2985}, {"id": 492, "name": "Peach", "freebase_id": "/m/0dj6p", "image_count": 86, "instance_count": 333}, {"id": 493, "name": "Coconut", "freebase_id": "/m/0djtd", "image_count": 81, "instance_count": 165}, {"id": 494, "name": "Seat belt", "freebase_id": "/m/0dkzw", "image_count": 311, "instance_count": 451}, {"id": 495, "name": "Raccoon", "freebase_id": "/m/0dq75", "image_count": 278, "instance_count": 336}, {"id": 496, "name": "Chisel", "freebase_id": "/m/0_dqb", "image_count": 20, "instance_count": 25}, {"id": 497, "name": "Fork", "freebase_id": "/m/0dt3t", "image_count": 885, "instance_count": 1442}, {"id": 498, "name": "Lamp", "freebase_id": "/m/0dtln", "image_count": 1420, "instance_count": 2912}, {"id": 499, "name": "Camera", "freebase_id": "/m/0dv5r", "image_count": 4726, "instance_count": 5933}, {"id": 500, "name": "Squash (Plant)", "freebase_id": "/m/0dv77", "image_count": 76, "instance_count": 157}, {"id": 501, "name": "Racket", "freebase_id": "/m/0dv9c", "image_count": 201, "instance_count": 236}, {"id": 502, "name": "Human face", "freebase_id": "/m/0dzct", "image_count": 286554, "instance_count": 762660}, {"id": 503, "name": "Human arm", "freebase_id": "/m/0dzf4", "image_count": 32749, "instance_count": 130607}, {"id": 504, "name": "Vegetable", "freebase_id": "/m/0f4s2w", "image_count": 1321, "instance_count": 4748}, {"id": 505, "name": "Diaper", "freebase_id": "/m/0f571", "image_count": 110, "instance_count": 146}, {"id": 506, "name": "Unicycle", "freebase_id": "/m/0f6nr", "image_count": 115, "instance_count": 159}, {"id": 507, "name": "Falcon", "freebase_id": "/m/0f6wt", "image_count": 1417, "instance_count": 1622}, {"id": 508, "name": "Chime", "freebase_id": "/m/0f8s22", "image_count": 21, "instance_count": 31}, {"id": 509, "name": "Snail", "freebase_id": "/m/0f9_l", "image_count": 552, "instance_count": 805}, {"id": 510, "name": "Shellfish", "freebase_id": "/m/0fbdv", "image_count": 245, "instance_count": 870}, {"id": 511, "name": "Cabbage", "freebase_id": "/m/0fbw6", "image_count": 98, "instance_count": 163}, {"id": 512, "name": "Carrot", "freebase_id": "/m/0fj52s", "image_count": 286, "instance_count": 699}, {"id": 513, "name": "Mango", "freebase_id": "/m/0fldg", "image_count": 57, "instance_count": 212}, {"id": 514, "name": "Jeans", "freebase_id": "/m/0fly7", "image_count": 24202, "instance_count": 56204}, {"id": 515, "name": "Flowerpot", "freebase_id": "/m/0fm3zh", "image_count": 4107, "instance_count": 9921}, {"id": 516, "name": "Pineapple", "freebase_id": "/m/0fp6w", "image_count": 230, "instance_count": 493}, {"id": 517, "name": "Drawer", "freebase_id": "/m/0fqfqc", "image_count": 617, "instance_count": 1892}, {"id": 518, "name": "Stool", "freebase_id": "/m/0fqt361", "image_count": 492, "instance_count": 1070}, {"id": 519, "name": "Envelope", "freebase_id": "/m/0frqm", "image_count": 116, "instance_count": 170}, {"id": 520, "name": "Cake", "freebase_id": "/m/0fszt", "image_count": 3076, "instance_count": 4100}, {"id": 521, "name": "Dragonfly", "freebase_id": "/m/0ft9s", "image_count": 1221, "instance_count": 1336}, {"id": 522, "name": "Common sunflower", "freebase_id": "/m/0ftb8", "image_count": 1419, "instance_count": 5807}, {"id": 523, "name": "Microwave oven", "freebase_id": "/m/0fx9l", "image_count": 337, "instance_count": 379}, {"id": 524, "name": "Honeycomb", "freebase_id": "/m/0fz0h", "image_count": 253, "instance_count": 272}, {"id": 525, "name": "Marine mammal", "freebase_id": "/m/0gd2v", "image_count": 873, "instance_count": 1428}, {"id": 526, "name": "Sea lion", "freebase_id": "/m/0gd36", "image_count": 737, "instance_count": 1406}, {"id": 527, "name": "Ladybug", "freebase_id": "/m/0gj37", "image_count": 519, "instance_count": 646}, {"id": 528, "name": "Shelf", "freebase_id": "/m/0gjbg72", "image_count": 2888, "instance_count": 9423}, {"id": 529, "name": "Watch", "freebase_id": "/m/0gjkl", "image_count": 1410, "instance_count": 1868}, {"id": 530, "name": "Candy", "freebase_id": "/m/0gm28", "image_count": 437, "instance_count": 1046}, {"id": 531, "name": "Salad", "freebase_id": "/m/0grw1", "image_count": 2241, "instance_count": 2576}, {"id": 532, "name": "Parrot", "freebase_id": "/m/0gv1x", "image_count": 1251, "instance_count": 1985}, {"id": 533, "name": "Handgun", "freebase_id": "/m/0gxl3", "image_count": 555, "instance_count": 710}, {"id": 534, "name": "Sparrow", "freebase_id": "/m/0h23m", "image_count": 1034, "instance_count": 1470}, {"id": 535, "name": "Van", "freebase_id": "/m/0h2r6", "image_count": 4349, "instance_count": 6200}, {"id": 536, "name": "Grinder", "freebase_id": "/m/0h8jyh6", "image_count": 10, "instance_count": 13}, {"id": 537, "name": "Spice rack", "freebase_id": "/m/0h8kx63", "image_count": 53, "instance_count": 105}, {"id": 538, "name": "Light bulb", "freebase_id": "/m/0h8l4fh", "image_count": 406, "instance_count": 1506}, {"id": 539, "name": "Corded phone", "freebase_id": "/m/0h8lkj8", "image_count": 344, "instance_count": 425}, {"id": 540, "name": "Sports uniform", "freebase_id": "/m/0h8mhzd", "image_count": 5416, "instance_count": 14241}, {"id": 541, "name": "Tennis racket", "freebase_id": "/m/0h8my_4", "image_count": 763, "instance_count": 884}, {"id": 542, "name": "Wall clock", "freebase_id": "/m/0h8mzrc", "image_count": 811, "instance_count": 983}, {"id": 543, "name": "Serving tray", "freebase_id": "/m/0h8n27j", "image_count": 99, "instance_count": 107}, {"id": 544, "name": "Kitchen & dining room table", "freebase_id": "/m/0h8n5zk", "image_count": 1495, "instance_count": 1597}, {"id": 545, "name": "Dog bed", "freebase_id": "/m/0h8n6f9", "image_count": 229, "instance_count": 253}, {"id": 546, "name": "Cake stand", "freebase_id": "/m/0h8n6ft", "image_count": 203, "instance_count": 276}, {"id": 547, "name": "Cat furniture", "freebase_id": "/m/0h8nm9j", "image_count": 180, "instance_count": 202}, {"id": 548, "name": "Bathroom accessory", "freebase_id": "/m/0h8nr_l", "image_count": 639, "instance_count": 2247}, {"id": 549, "name": "Facial tissue holder", "freebase_id": "/m/0h8nsvg", "image_count": 18, "instance_count": 19}, {"id": 550, "name": "Pressure cooker", "freebase_id": "/m/0h8ntjv", "image_count": 13, "instance_count": 14}, {"id": 551, "name": "Kitchen appliance", "freebase_id": "/m/0h99cwc", "image_count": 928, "instance_count": 3552}, {"id": 552, "name": "Tire", "freebase_id": "/m/0h9mv", "image_count": 23939, "instance_count": 83517}, {"id": 553, "name": "Ruler", "freebase_id": "/m/0hdln", "image_count": 230, "instance_count": 245}, {"id": 554, "name": "Luggage and bags", "freebase_id": "/m/0hf58v5", "image_count": 801, "instance_count": 1774}, {"id": 555, "name": "Microphone", "freebase_id": "/m/0hg7b", "image_count": 19830, "instance_count": 26336}, {"id": 556, "name": "Broccoli", "freebase_id": "/m/0hkxq", "image_count": 222, "instance_count": 712}, {"id": 557, "name": "Umbrella", "freebase_id": "/m/0hnnb", "image_count": 2273, "instance_count": 4705}, {"id": 558, "name": "Pastry", "freebase_id": "/m/0hnyx", "image_count": 642, "instance_count": 2094}, {"id": 559, "name": "Grapefruit", "freebase_id": "/m/0hqkz", "image_count": 297, "instance_count": 768}, {"id": 560, "name": "Band-aid", "freebase_id": "/m/0j496", "image_count": 20, "instance_count": 36}, {"id": 561, "name": "Animal", "freebase_id": "/m/0jbk", "image_count": 7214, "instance_count": 12475}, {"id": 562, "name": "Bell pepper", "freebase_id": "/m/0jg57", "image_count": 198, "instance_count": 490}, {"id": 563, "name": "Turkey", "freebase_id": "/m/0jly1", "image_count": 290, "instance_count": 570}, {"id": 564, "name": "Lily", "freebase_id": "/m/0jqgx", "image_count": 863, "instance_count": 1678}, {"id": 565, "name": "Pomegranate", "freebase_id": "/m/0jwn_", "image_count": 159, "instance_count": 343}, {"id": 566, "name": "Doughnut", "freebase_id": "/m/0jy4k", "image_count": 241, "instance_count": 552}, {"id": 567, "name": "Glasses", "freebase_id": "/m/0jyfg", "image_count": 38197, "instance_count": 50948}, {"id": 568, "name": "Human nose", "freebase_id": "/m/0k0pj", "image_count": 24508, "instance_count": 43663}, {"id": 569, "name": "Pen", "freebase_id": "/m/0k1tl", "image_count": 926, "instance_count": 1492}, {"id": 570, "name": "Ant", "freebase_id": "/m/0_k2", "image_count": 327, "instance_count": 784}, {"id": 571, "name": "Car", "freebase_id": "/m/0k4j", "image_count": 63478, "instance_count": 164003}, {"id": 572, "name": "Aircraft", "freebase_id": "/m/0k5j", "image_count": 1039, "instance_count": 1615}, {"id": 573, "name": "Human hand", "freebase_id": "/m/0k65p", "image_count": 18436, "instance_count": 51215}, {"id": 574, "name": "Skunk", "freebase_id": "/m/0km7z", "image_count": 36, "instance_count": 51}, {"id": 575, "name": "Teddy bear", "freebase_id": "/m/0kmg4", "image_count": 926, "instance_count": 1401}, {"id": 576, "name": "Watermelon", "freebase_id": "/m/0kpqd", "image_count": 255, "instance_count": 551}, {"id": 577, "name": "Cantaloupe", "freebase_id": "/m/0kpt_", "image_count": 69, "instance_count": 114}, {"id": 578, "name": "Dishwasher", "freebase_id": "/m/0ky7b", "image_count": 72, "instance_count": 74}, {"id": 579, "name": "Flute", "freebase_id": "/m/0l14j_", "image_count": 228, "instance_count": 340}, {"id": 580, "name": "Balance beam", "freebase_id": "/m/0l3ms", "image_count": 247, "instance_count": 309}, {"id": 581, "name": "Sandwich", "freebase_id": "/m/0l515", "image_count": 576, "instance_count": 831}, {"id": 582, "name": "Shrimp", "freebase_id": "/m/0ll1f78", "image_count": 373, "instance_count": 1344}, {"id": 583, "name": "Sewing machine", "freebase_id": "/m/0llzx", "image_count": 351, "instance_count": 425}, {"id": 584, "name": "Binoculars", "freebase_id": "/m/0lt4_", "image_count": 101, "instance_count": 115}, {"id": 585, "name": "Rays and skates", "freebase_id": "/m/0m53l", "image_count": 319, "instance_count": 479}, {"id": 586, "name": "Ipod", "freebase_id": "/m/0mcx2", "image_count": 403, "instance_count": 570}, {"id": 587, "name": "Accordion", "freebase_id": "/m/0mkg", "image_count": 774, "instance_count": 862}, {"id": 588, "name": "Willow", "freebase_id": "/m/0mw_6", "image_count": 26, "instance_count": 43}, {"id": 589, "name": "Crab", "freebase_id": "/m/0n28_", "image_count": 616, "instance_count": 828}, {"id": 590, "name": "Crown", "freebase_id": "/m/0nl46", "image_count": 514, "instance_count": 612}, {"id": 591, "name": "Seahorse", "freebase_id": "/m/0nybt", "image_count": 219, "instance_count": 284}, {"id": 592, "name": "Perfume", "freebase_id": "/m/0p833", "image_count": 193, "instance_count": 304}, {"id": 593, "name": "Alpaca", "freebase_id": "/m/0pcr", "image_count": 351, "instance_count": 593}, {"id": 594, "name": "Taxi", "freebase_id": "/m/0pg52", "image_count": 951, "instance_count": 2491}, {"id": 595, "name": "Canoe", "freebase_id": "/m/0ph39", "image_count": 1439, "instance_count": 2881}, {"id": 596, "name": "Remote control", "freebase_id": "/m/0qjjc", "image_count": 183, "instance_count": 221}, {"id": 597, "name": "Wheelchair", "freebase_id": "/m/0qmmr", "image_count": 737, "instance_count": 1181}, {"id": 598, "name": "Rugby ball", "freebase_id": "/m/0wdt60w", "image_count": 225, "instance_count": 244}, {"id": 599, "name": "Armadillo", "freebase_id": "/m/0xfy", "image_count": 42, "instance_count": 54}, {"id": 600, "name": "Maracas", "freebase_id": "/m/0xzly", "image_count": 3, "instance_count": 5}, {"id": 601, "name": "Helmet", "freebase_id": "/m/0zvk5", "image_count": 6360, "instance_count": 11798}]
+# fmt: on

approach/ovod/APE/ape/data/datasets/pascal_voc_external.py

@@ -0,0 +1,1217 @@
+import os
+
+from detectron2.data import DatasetCatalog, MetadataCatalog
+from detectron2.data.datasets import load_sem_seg
+
+from .coco import custom_register_coco_instances
+
+PASCAL_CTX_59_CATEGORIES = [
+    {"color": [180, 120, 120], "id": 0, "isthing": 0, "name": "aeroplane"},
+    {"color": [6, 230, 230], "id": 1, "isthing": 0, "name": "bag"},
+    {"color": [80, 50, 50], "id": 2, "isthing": 0, "name": "bed"},
+    {"color": [4, 200, 3], "id": 3, "isthing": 0, "name": "bedclothes"},
+    {"color": [120, 120, 80], "id": 4, "isthing": 0, "name": "bench"},
+    {"color": [140, 140, 140], "id": 5, "isthing": 0, "name": "bicycle"},
+    {"color": [204, 5, 255], "id": 6, "isthing": 0, "name": "bird"},
+    {"color": [230, 230, 230], "id": 7, "isthing": 0, "name": "boat"},
+    {"color": [4, 250, 7], "id": 8, "isthing": 0, "name": "book"},
+    {"color": [224, 5, 255], "id": 9, "isthing": 0, "name": "bottle"},
+    {"color": [235, 255, 7], "id": 10, "isthing": 0, "name": "building"},
+    {"color": [150, 5, 61], "id": 11, "isthing": 0, "name": "bus"},
+    {"color": [120, 120, 70], "id": 12, "isthing": 0, "name": "cabinet"},
+    {"color": [8, 255, 51], "id": 13, "isthing": 0, "name": "car"},
+    {"color": [255, 6, 82], "id": 14, "isthing": 0, "name": "cat"},
+    {"color": [143, 255, 140], "id": 15, "isthing": 0, "name": "ceiling"},
+    {"color": [204, 255, 4], "id": 16, "isthing": 0, "name": "chair"},
+    {"color": [255, 51, 7], "id": 17, "isthing": 0, "name": "cloth"},
+    {"color": [204, 70, 3], "id": 18, "isthing": 0, "name": "computer"},
+    {"color": [0, 102, 200], "id": 19, "isthing": 0, "name": "cow"},
+    {"color": [61, 230, 250], "id": 20, "isthing": 0, "name": "cup"},
+    {"color": [255, 6, 51], "id": 21, "isthing": 0, "name": "curtain"},
+    {"color": [11, 102, 255], "id": 22, "isthing": 0, "name": "dog"},
+    {"color": [255, 7, 71], "id": 23, "isthing": 0, "name": "door"},
+    {"color": [255, 9, 224], "id": 24, "isthing": 0, "name": "fence"},
+    {"color": [9, 7, 230], "id": 25, "isthing": 0, "name": "floor"},
+    {"color": [220, 220, 220], "id": 26, "isthing": 0, "name": "flower"},
+    {"color": [255, 9, 92], "id": 27, "isthing": 0, "name": "food"},
+    {"color": [112, 9, 255], "id": 28, "isthing": 0, "name": "grass"},
+    {"color": [8, 255, 214], "id": 29, "isthing": 0, "name": "ground"},
+    {"color": [7, 255, 224], "id": 30, "isthing": 0, "name": "horse"},
+    {"color": [255, 184, 6], "id": 31, "isthing": 0, "name": "keyboard"},
+    {"color": [10, 255, 71], "id": 32, "isthing": 0, "name": "light"},
+    {"color": [255, 41, 10], "id": 33, "isthing": 0, "name": "motorbike"},
+    {"color": [7, 255, 255], "id": 34, "isthing": 0, "name": "mountain"},
+    {"color": [224, 255, 8], "id": 35, "isthing": 0, "name": "mouse"},
+    {"color": [102, 8, 255], "id": 36, "isthing": 0, "name": "person"},
+    {"color": [255, 61, 6], "id": 37, "isthing": 0, "name": "plate"},
+    {"color": [255, 194, 7], "id": 38, "isthing": 0, "name": "platform"},
+    {"color": [255, 122, 8], "id": 39, "isthing": 0, "name": "pottedplant"},
+    {"color": [0, 255, 20], "id": 40, "isthing": 0, "name": "road"},
+    {"color": [255, 8, 41], "id": 41, "isthing": 0, "name": "rock"},
+    {"color": [255, 5, 153], "id": 42, "isthing": 0, "name": "sheep"},
+    {"color": [6, 51, 255], "id": 43, "isthing": 0, "name": "shelves"},
+    {"color": [235, 12, 255], "id": 44, "isthing": 0, "name": "sidewalk"},
+    {"color": [160, 150, 20], "id": 45, "isthing": 0, "name": "sign"},
+    {"color": [0, 163, 255], "id": 46, "isthing": 0, "name": "sky"},
+    {"color": [140, 140, 140], "id": 47, "isthing": 0, "name": "snow"},
+    {"color": [250, 10, 15], "id": 48, "isthing": 0, "name": "sofa"},
+    {"color": [20, 255, 0], "id": 49, "isthing": 0, "name": "diningtable"},
+    {"color": [31, 255, 0], "id": 50, "isthing": 0, "name": "track"},
+    {"color": [255, 31, 0], "id": 51, "isthing": 0, "name": "train"},
+    {"color": [255, 224, 0], "id": 52, "isthing": 0, "name": "tree"},
+    {"color": [153, 255, 0], "id": 53, "isthing": 0, "name": "truck"},
+    {"color": [0, 0, 255], "id": 54, "isthing": 0, "name": "tvmonitor"},
+    {"color": [255, 71, 0], "id": 55, "isthing": 0, "name": "wall"},
+    {"color": [0, 235, 255], "id": 56, "isthing": 0, "name": "water"},
+    {"color": [0, 173, 255], "id": 57, "isthing": 0, "name": "window"},
+    {"color": [31, 0, 255], "id": 58, "isthing": 0, "name": "wood"},
+]
+
+PASCAL_CTX_459_CATEGORIES = [
+    {"color": [120, 120, 120], "id": 0, "isthing": 0, "name": "accordion"},
+    {"color": [180, 120, 120], "id": 1, "isthing": 0, "name": "aeroplane"},
+    {"color": [6, 230, 230], "id": 2, "isthing": 0, "name": "air conditioner"},
+    {"color": [80, 50, 50], "id": 3, "isthing": 0, "name": "antenna"},
+    {"color": [4, 200, 3], "id": 4, "isthing": 0, "name": "artillery"},
+    {"color": [120, 120, 80], "id": 5, "isthing": 0, "name": "ashtray"},
+    {"color": [140, 140, 140], "id": 6, "isthing": 0, "name": "atrium"},
+    {"color": [204, 5, 255], "id": 7, "isthing": 0, "name": "baby carriage"},
+    {"color": [230, 230, 230], "id": 8, "isthing": 0, "name": "bag"},
+    {"color": [4, 250, 7], "id": 9, "isthing": 0, "name": "ball"},
+    {"color": [224, 5, 255], "id": 10, "isthing": 0, "name": "balloon"},
+    {"color": [235, 255, 7], "id": 11, "isthing": 0, "name": "bamboo weaving"},
+    {"color": [150, 5, 61], "id": 12, "isthing": 0, "name": "barrel"},
+    {"color": [120, 120, 70], "id": 13, "isthing": 0, "name": "baseball bat"},
+    {"color": [8, 255, 51], "id": 14, "isthing": 0, "name": "basket"},
+    {"color": [255, 6, 82], "id": 15, "isthing": 0, "name": "basketball backboard"},
+    {"color": [143, 255, 140], "id": 16, "isthing": 0, "name": "bathtub"},
+    {"color": [204, 255, 4], "id": 17, "isthing": 0, "name": "bed"},
+    {"color": [255, 51, 7], "id": 18, "isthing": 0, "name": "bedclothes"},
+    {"color": [204, 70, 3], "id": 19, "isthing": 0, "name": "beer"},
+    {"color": [0, 102, 200], "id": 20, "isthing": 0, "name": "bell"},
+    {"color": [61, 230, 250], "id": 21, "isthing": 0, "name": "bench"},
+    {"color": [255, 6, 51], "id": 22, "isthing": 0, "name": "bicycle"},
+    {"color": [11, 102, 255], "id": 23, "isthing": 0, "name": "binoculars"},
+    {"color": [255, 7, 71], "id": 24, "isthing": 0, "name": "bird"},
+    {"color": [255, 9, 224], "id": 25, "isthing": 0, "name": "bird cage"},
+    {"color": [9, 7, 230], "id": 26, "isthing": 0, "name": "bird feeder"},
+    {"color": [220, 220, 220], "id": 27, "isthing": 0, "name": "bird nest"},
+    {"color": [255, 9, 92], "id": 28, "isthing": 0, "name": "blackboard"},
+    {"color": [112, 9, 255], "id": 29, "isthing": 0, "name": "board"},
+    {"color": [8, 255, 214], "id": 30, "isthing": 0, "name": "boat"},
+    {"color": [7, 255, 224], "id": 31, "isthing": 0, "name": "bone"},
+    {"color": [255, 184, 6], "id": 32, "isthing": 0, "name": "book"},
+    {"color": [10, 255, 71], "id": 33, "isthing": 0, "name": "bottle"},
+    {"color": [255, 41, 10], "id": 34, "isthing": 0, "name": "bottle opener"},
+    {"color": [7, 255, 255], "id": 35, "isthing": 0, "name": "bowl"},
+    {"color": [224, 255, 8], "id": 36, "isthing": 0, "name": "box"},
+    {"color": [102, 8, 255], "id": 37, "isthing": 0, "name": "bracelet"},
+    {"color": [255, 61, 6], "id": 38, "isthing": 0, "name": "brick"},
+    {"color": [255, 194, 7], "id": 39, "isthing": 0, "name": "bridge"},
+    {"color": [255, 122, 8], "id": 40, "isthing": 0, "name": "broom"},
+    {"color": [0, 255, 20], "id": 41, "isthing": 0, "name": "brush"},
+    {"color": [255, 8, 41], "id": 42, "isthing": 0, "name": "bucket"},
+    {"color": [255, 5, 153], "id": 43, "isthing": 0, "name": "building"},
+    {"color": [6, 51, 255], "id": 44, "isthing": 0, "name": "bus"},
+    {"color": [235, 12, 255], "id": 45, "isthing": 0, "name": "cabinet"},
+    {"color": [160, 150, 20], "id": 46, "isthing": 0, "name": "cabinet door"},
+    {"color": [0, 163, 255], "id": 47, "isthing": 0, "name": "cage"},
+    {"color": [140, 140, 140], "id": 48, "isthing": 0, "name": "cake"},
+    {"color": [250, 10, 15], "id": 49, "isthing": 0, "name": "calculator"},
+    {"color": [20, 255, 0], "id": 50, "isthing": 0, "name": "calendar"},
+    {"color": [31, 255, 0], "id": 51, "isthing": 0, "name": "camel"},
+    {"color": [255, 31, 0], "id": 52, "isthing": 0, "name": "camera"},
+    {"color": [255, 224, 0], "id": 53, "isthing": 0, "name": "camera lens"},
+    {"color": [153, 255, 0], "id": 54, "isthing": 0, "name": "can"},
+    {"color": [0, 0, 255], "id": 55, "isthing": 0, "name": "candle"},
+    {"color": [255, 71, 0], "id": 56, "isthing": 0, "name": "candle holder"},
+    {"color": [0, 235, 255], "id": 57, "isthing": 0, "name": "cap"},
+    {"color": [0, 173, 255], "id": 58, "isthing": 0, "name": "car"},
+    {"color": [31, 0, 255], "id": 59, "isthing": 0, "name": "card"},
+    {"color": [120, 120, 120], "id": 60, "isthing": 0, "name": "cart"},
+    {"color": [180, 120, 120], "id": 61, "isthing": 0, "name": "case"},
+    {"color": [6, 230, 230], "id": 62, "isthing": 0, "name": "casette recorder"},
+    {"color": [80, 50, 50], "id": 63, "isthing": 0, "name": "cash register"},
+    {"color": [4, 200, 3], "id": 64, "isthing": 0, "name": "cat"},
+    {"color": [120, 120, 80], "id": 65, "isthing": 0, "name": "cd"},
+    {"color": [140, 140, 140], "id": 66, "isthing": 0, "name": "cd player"},
+    {"color": [204, 5, 255], "id": 67, "isthing": 0, "name": "ceiling"},
+    {"color": [230, 230, 230], "id": 68, "isthing": 0, "name": "cell phone"},
+    {"color": [4, 250, 7], "id": 69, "isthing": 0, "name": "cello"},
+    {"color": [224, 5, 255], "id": 70, "isthing": 0, "name": "chain"},
+    {"color": [235, 255, 7], "id": 71, "isthing": 0, "name": "chair"},
+    {"color": [150, 5, 61], "id": 72, "isthing": 0, "name": "chessboard"},
+    {"color": [120, 120, 70], "id": 73, "isthing": 0, "name": "chicken"},
+    {"color": [8, 255, 51], "id": 74, "isthing": 0, "name": "chopstick"},
+    {"color": [255, 6, 82], "id": 75, "isthing": 0, "name": "clip"},
+    {"color": [143, 255, 140], "id": 76, "isthing": 0, "name": "clippers"},
+    {"color": [204, 255, 4], "id": 77, "isthing": 0, "name": "clock"},
+    {"color": [255, 51, 7], "id": 78, "isthing": 0, "name": "closet"},
+    {"color": [204, 70, 3], "id": 79, "isthing": 0, "name": "cloth"},
+    {"color": [0, 102, 200], "id": 80, "isthing": 0, "name": "clothes tree"},
+    {"color": [61, 230, 250], "id": 81, "isthing": 0, "name": "coffee"},
+    {"color": [255, 6, 51], "id": 82, "isthing": 0, "name": "coffee machine"},
+    {"color": [11, 102, 255], "id": 83, "isthing": 0, "name": "comb"},
+    {"color": [255, 7, 71], "id": 84, "isthing": 0, "name": "computer"},
+    {"color": [255, 9, 224], "id": 85, "isthing": 0, "name": "concrete"},
+    {"color": [9, 7, 230], "id": 86, "isthing": 0, "name": "cone"},
+    {"color": [220, 220, 220], "id": 87, "isthing": 0, "name": "container"},
+    {"color": [255, 9, 92], "id": 88, "isthing": 0, "name": "control booth"},
+    {"color": [112, 9, 255], "id": 89, "isthing": 0, "name": "controller"},
+    {"color": [8, 255, 214], "id": 90, "isthing": 0, "name": "cooker"},
+    {"color": [7, 255, 224], "id": 91, "isthing": 0, "name": "copying machine"},
+    {"color": [255, 184, 6], "id": 92, "isthing": 0, "name": "coral"},
+    {"color": [10, 255, 71], "id": 93, "isthing": 0, "name": "cork"},
+    {"color": [255, 41, 10], "id": 94, "isthing": 0, "name": "corkscrew"},
+    {"color": [7, 255, 255], "id": 95, "isthing": 0, "name": "counter"},
+    {"color": [224, 255, 8], "id": 96, "isthing": 0, "name": "court"},
+    {"color": [102, 8, 255], "id": 97, "isthing": 0, "name": "cow"},
+    {"color": [255, 61, 6], "id": 98, "isthing": 0, "name": "crabstick"},
+    {"color": [255, 194, 7], "id": 99, "isthing": 0, "name": "crane"},
+    {"color": [255, 122, 8], "id": 100, "isthing": 0, "name": "crate"},
+    {"color": [0, 255, 20], "id": 101, "isthing": 0, "name": "cross"},
+    {"color": [255, 8, 41], "id": 102, "isthing": 0, "name": "crutch"},
+    {"color": [255, 5, 153], "id": 103, "isthing": 0, "name": "cup"},
+    {"color": [6, 51, 255], "id": 104, "isthing": 0, "name": "curtain"},
+    {"color": [235, 12, 255], "id": 105, "isthing": 0, "name": "cushion"},
+    {"color": [160, 150, 20], "id": 106, "isthing": 0, "name": "cutting board"},
+    {"color": [0, 163, 255], "id": 107, "isthing": 0, "name": "dais"},
+    {"color": [140, 140, 140], "id": 108, "isthing": 0, "name": "disc"},
+    {"color": [250, 10, 15], "id": 109, "isthing": 0, "name": "disc case"},
+    {"color": [20, 255, 0], "id": 110, "isthing": 0, "name": "dishwasher"},
+    {"color": [31, 255, 0], "id": 111, "isthing": 0, "name": "dock"},
+    {"color": [255, 31, 0], "id": 112, "isthing": 0, "name": "dog"},
+    {"color": [255, 224, 0], "id": 113, "isthing": 0, "name": "dolphin"},
+    {"color": [153, 255, 0], "id": 114, "isthing": 0, "name": "door"},
+    {"color": [0, 0, 255], "id": 115, "isthing": 0, "name": "drainer"},
+    {"color": [255, 71, 0], "id": 116, "isthing": 0, "name": "dray"},
+    {"color": [0, 235, 255], "id": 117, "isthing": 0, "name": "drink dispenser"},
+    {"color": [0, 173, 255], "id": 118, "isthing": 0, "name": "drinking machine"},
+    {"color": [31, 0, 255], "id": 119, "isthing": 0, "name": "drop"},
+    {"color": [120, 120, 120], "id": 120, "isthing": 0, "name": "drug"},
+    {"color": [180, 120, 120], "id": 121, "isthing": 0, "name": "drum"},
+    {"color": [6, 230, 230], "id": 122, "isthing": 0, "name": "drum kit"},
+    {"color": [80, 50, 50], "id": 123, "isthing": 0, "name": "duck"},
+    {"color": [4, 200, 3], "id": 124, "isthing": 0, "name": "dumbbell"},
+    {"color": [120, 120, 80], "id": 125, "isthing": 0, "name": "earphone"},
+    {"color": [140, 140, 140], "id": 126, "isthing": 0, "name": "earrings"},
+    {"color": [204, 5, 255], "id": 127, "isthing": 0, "name": "egg"},
+    {"color": [230, 230, 230], "id": 128, "isthing": 0, "name": "electric fan"},
+    {"color": [4, 250, 7], "id": 129, "isthing": 0, "name": "electric iron"},
+    {"color": [224, 5, 255], "id": 130, "isthing": 0, "name": "electric pot"},
+    {"color": [235, 255, 7], "id": 131, "isthing": 0, "name": "electric saw"},
+    {"color": [150, 5, 61], "id": 132, "isthing": 0, "name": "electronic keyboard"},
+    {"color": [120, 120, 70], "id": 133, "isthing": 0, "name": "engine"},
+    {"color": [8, 255, 51], "id": 134, "isthing": 0, "name": "envelope"},
+    {"color": [255, 6, 82], "id": 135, "isthing": 0, "name": "equipment"},
+    {"color": [143, 255, 140], "id": 136, "isthing": 0, "name": "escalator"},
+    {"color": [204, 255, 4], "id": 137, "isthing": 0, "name": "exhibition booth"},
+    {"color": [255, 51, 7], "id": 138, "isthing": 0, "name": "extinguisher"},
+    {"color": [204, 70, 3], "id": 139, "isthing": 0, "name": "eyeglass"},
+    {"color": [0, 102, 200], "id": 140, "isthing": 0, "name": "fan"},
+    {"color": [61, 230, 250], "id": 141, "isthing": 0, "name": "faucet"},
+    {"color": [255, 6, 51], "id": 142, "isthing": 0, "name": "fax machine"},
+    {"color": [11, 102, 255], "id": 143, "isthing": 0, "name": "fence"},
+    {"color": [255, 7, 71], "id": 144, "isthing": 0, "name": "ferris wheel"},
+    {"color": [255, 9, 224], "id": 145, "isthing": 0, "name": "fire extinguisher"},
+    {"color": [9, 7, 230], "id": 146, "isthing": 0, "name": "fire hydrant"},
+    {"color": [220, 220, 220], "id": 147, "isthing": 0, "name": "fire place"},
+    {"color": [255, 9, 92], "id": 148, "isthing": 0, "name": "fish"},
+    {"color": [112, 9, 255], "id": 149, "isthing": 0, "name": "fish tank"},
+    {"color": [8, 255, 214], "id": 150, "isthing": 0, "name": "fishbowl"},
+    {"color": [7, 255, 224], "id": 151, "isthing": 0, "name": "fishing net"},
+    {"color": [255, 184, 6], "id": 152, "isthing": 0, "name": "fishing pole"},
+    {"color": [10, 255, 71], "id": 153, "isthing": 0, "name": "flag"},
+    {"color": [255, 41, 10], "id": 154, "isthing": 0, "name": "flagstaff"},
+    {"color": [7, 255, 255], "id": 155, "isthing": 0, "name": "flame"},
+    {"color": [224, 255, 8], "id": 156, "isthing": 0, "name": "flashlight"},
+    {"color": [102, 8, 255], "id": 157, "isthing": 0, "name": "floor"},
+    {"color": [255, 61, 6], "id": 158, "isthing": 0, "name": "flower"},
+    {"color": [255, 194, 7], "id": 159, "isthing": 0, "name": "fly"},
+    {"color": [255, 122, 8], "id": 160, "isthing": 0, "name": "foam"},
+    {"color": [0, 255, 20], "id": 161, "isthing": 0, "name": "food"},
+    {"color": [255, 8, 41], "id": 162, "isthing": 0, "name": "footbridge"},
+    {"color": [255, 5, 153], "id": 163, "isthing": 0, "name": "forceps"},
+    {"color": [6, 51, 255], "id": 164, "isthing": 0, "name": "fork"},
+    {"color": [235, 12, 255], "id": 165, "isthing": 0, "name": "forklift"},
+    {"color": [160, 150, 20], "id": 166, "isthing": 0, "name": "fountain"},
+    {"color": [0, 163, 255], "id": 167, "isthing": 0, "name": "fox"},
+    {"color": [140, 140, 140], "id": 168, "isthing": 0, "name": "frame"},
+    {"color": [250, 10, 15], "id": 169, "isthing": 0, "name": "fridge"},
+    {"color": [20, 255, 0], "id": 170, "isthing": 0, "name": "frog"},
+    {"color": [31, 255, 0], "id": 171, "isthing": 0, "name": "fruit"},
+    {"color": [255, 31, 0], "id": 172, "isthing": 0, "name": "funnel"},
+    {"color": [255, 224, 0], "id": 173, "isthing": 0, "name": "furnace"},
+    {"color": [153, 255, 0], "id": 174, "isthing": 0, "name": "game controller"},
+    {"color": [0, 0, 255], "id": 175, "isthing": 0, "name": "game machine"},
+    {"color": [255, 71, 0], "id": 176, "isthing": 0, "name": "gas cylinder"},
+    {"color": [0, 235, 255], "id": 177, "isthing": 0, "name": "gas hood"},
+    {"color": [0, 173, 255], "id": 178, "isthing": 0, "name": "gas stove"},
+    {"color": [31, 0, 255], "id": 179, "isthing": 0, "name": "gift box"},
+    {"color": [120, 120, 120], "id": 180, "isthing": 0, "name": "glass"},
+    {"color": [180, 120, 120], "id": 181, "isthing": 0, "name": "glass marble"},
+    {"color": [6, 230, 230], "id": 182, "isthing": 0, "name": "globe"},
+    {"color": [80, 50, 50], "id": 183, "isthing": 0, "name": "glove"},
+    {"color": [4, 200, 3], "id": 184, "isthing": 0, "name": "goal"},
+    {"color": [120, 120, 80], "id": 185, "isthing": 0, "name": "grandstand"},
+    {"color": [140, 140, 140], "id": 186, "isthing": 0, "name": "grass"},
+    {"color": [204, 5, 255], "id": 187, "isthing": 0, "name": "gravestone"},
+    {"color": [230, 230, 230], "id": 188, "isthing": 0, "name": "ground"},
+    {"color": [4, 250, 7], "id": 189, "isthing": 0, "name": "guardrail"},
+    {"color": [224, 5, 255], "id": 190, "isthing": 0, "name": "guitar"},
+    {"color": [235, 255, 7], "id": 191, "isthing": 0, "name": "gun"},
+    {"color": [150, 5, 61], "id": 192, "isthing": 0, "name": "hammer"},
+    {"color": [120, 120, 70], "id": 193, "isthing": 0, "name": "hand cart"},
+    {"color": [8, 255, 51], "id": 194, "isthing": 0, "name": "handle"},
+    {"color": [255, 6, 82], "id": 195, "isthing": 0, "name": "handrail"},
+    {"color": [143, 255, 140], "id": 196, "isthing": 0, "name": "hanger"},
+    {"color": [204, 255, 4], "id": 197, "isthing": 0, "name": "hard disk drive"},
+    {"color": [255, 51, 7], "id": 198, "isthing": 0, "name": "hat"},
+    {"color": [204, 70, 3], "id": 199, "isthing": 0, "name": "hay"},
+    {"color": [0, 102, 200], "id": 200, "isthing": 0, "name": "headphone"},
+    {"color": [61, 230, 250], "id": 201, "isthing": 0, "name": "heater"},
+    {"color": [255, 6, 51], "id": 202, "isthing": 0, "name": "helicopter"},
+    {"color": [11, 102, 255], "id": 203, "isthing": 0, "name": "helmet"},
+    {"color": [255, 7, 71], "id": 204, "isthing": 0, "name": "holder"},
+    {"color": [255, 9, 224], "id": 205, "isthing": 0, "name": "hook"},
+    {"color": [9, 7, 230], "id": 206, "isthing": 0, "name": "horse"},
+    {"color": [220, 220, 220], "id": 207, "isthing": 0, "name": "horse-drawn carriage"},
+    {"color": [255, 9, 92], "id": 208, "isthing": 0, "name": "hot-air balloon"},
+    {"color": [112, 9, 255], "id": 209, "isthing": 0, "name": "hydrovalve"},
+    {"color": [8, 255, 214], "id": 210, "isthing": 0, "name": "ice"},
+    {"color": [7, 255, 224], "id": 211, "isthing": 0, "name": "inflator pump"},
+    {"color": [255, 184, 6], "id": 212, "isthing": 0, "name": "ipod"},
+    {"color": [10, 255, 71], "id": 213, "isthing": 0, "name": "iron"},
+    {"color": [255, 41, 10], "id": 214, "isthing": 0, "name": "ironing board"},
+    {"color": [7, 255, 255], "id": 215, "isthing": 0, "name": "jar"},
+    {"color": [224, 255, 8], "id": 216, "isthing": 0, "name": "kart"},
+    {"color": [102, 8, 255], "id": 217, "isthing": 0, "name": "kettle"},
+    {"color": [255, 61, 6], "id": 218, "isthing": 0, "name": "key"},
+    {"color": [255, 194, 7], "id": 219, "isthing": 0, "name": "keyboard"},
+    {"color": [255, 122, 8], "id": 220, "isthing": 0, "name": "kitchen range"},
+    {"color": [0, 255, 20], "id": 221, "isthing": 0, "name": "kite"},
+    {"color": [255, 8, 41], "id": 222, "isthing": 0, "name": "knife"},
+    {"color": [255, 5, 153], "id": 223, "isthing": 0, "name": "knife block"},
+    {"color": [6, 51, 255], "id": 224, "isthing": 0, "name": "ladder"},
+    {"color": [235, 12, 255], "id": 225, "isthing": 0, "name": "ladder truck"},
+    {"color": [160, 150, 20], "id": 226, "isthing": 0, "name": "ladle"},
+    {"color": [0, 163, 255], "id": 227, "isthing": 0, "name": "laptop"},
+    {"color": [140, 140, 140], "id": 228, "isthing": 0, "name": "leaves"},
+    {"color": [250, 10, 15], "id": 229, "isthing": 0, "name": "lid"},
+    {"color": [20, 255, 0], "id": 230, "isthing": 0, "name": "life buoy"},
+    {"color": [31, 255, 0], "id": 231, "isthing": 0, "name": "light"},
+    {"color": [255, 31, 0], "id": 232, "isthing": 0, "name": "light bulb"},
+    {"color": [255, 224, 0], "id": 233, "isthing": 0, "name": "lighter"},
+    {"color": [153, 255, 0], "id": 234, "isthing": 0, "name": "line"},
+    {"color": [0, 0, 255], "id": 235, "isthing": 0, "name": "lion"},
+    {"color": [255, 71, 0], "id": 236, "isthing": 0, "name": "lobster"},
+    {"color": [0, 235, 255], "id": 237, "isthing": 0, "name": "lock"},
+    {"color": [0, 173, 255], "id": 238, "isthing": 0, "name": "machine"},
+    {"color": [31, 0, 255], "id": 239, "isthing": 0, "name": "mailbox"},
+    {"color": [120, 120, 120], "id": 240, "isthing": 0, "name": "mannequin"},
+    {"color": [180, 120, 120], "id": 241, "isthing": 0, "name": "map"},
+    {"color": [6, 230, 230], "id": 242, "isthing": 0, "name": "mask"},
+    {"color": [80, 50, 50], "id": 243, "isthing": 0, "name": "mat"},
+    {"color": [4, 200, 3], "id": 244, "isthing": 0, "name": "match book"},
+    {"color": [120, 120, 80], "id": 245, "isthing": 0, "name": "mattress"},
+    {"color": [140, 140, 140], "id": 246, "isthing": 0, "name": "menu"},
+    {"color": [204, 5, 255], "id": 247, "isthing": 0, "name": "metal"},
+    {"color": [230, 230, 230], "id": 248, "isthing": 0, "name": "meter box"},
+    {"color": [4, 250, 7], "id": 249, "isthing": 0, "name": "microphone"},
+    {"color": [224, 5, 255], "id": 250, "isthing": 0, "name": "microwave"},
+    {"color": [235, 255, 7], "id": 251, "isthing": 0, "name": "mirror"},
+    {"color": [150, 5, 61], "id": 252, "isthing": 0, "name": "missile"},
+    {"color": [120, 120, 70], "id": 253, "isthing": 0, "name": "model"},
+    {"color": [8, 255, 51], "id": 254, "isthing": 0, "name": "money"},
+    {"color": [255, 6, 82], "id": 255, "isthing": 0, "name": "monkey"},
+    {"color": [143, 255, 140], "id": 256, "isthing": 0, "name": "mop"},
+    {"color": [204, 255, 4], "id": 257, "isthing": 0, "name": "motorbike"},
+    {"color": [255, 51, 7], "id": 258, "isthing": 0, "name": "mountain"},
+    {"color": [204, 70, 3], "id": 259, "isthing": 0, "name": "mouse"},
+    {"color": [0, 102, 200], "id": 260, "isthing": 0, "name": "mouse pad"},
+    {"color": [61, 230, 250], "id": 261, "isthing": 0, "name": "musical instrument"},
+    {"color": [255, 6, 51], "id": 262, "isthing": 0, "name": "napkin"},
+    {"color": [11, 102, 255], "id": 263, "isthing": 0, "name": "net"},
+    {"color": [255, 7, 71], "id": 264, "isthing": 0, "name": "newspaper"},
+    {"color": [255, 9, 224], "id": 265, "isthing": 0, "name": "oar"},
+    {"color": [9, 7, 230], "id": 266, "isthing": 0, "name": "ornament"},
+    {"color": [220, 220, 220], "id": 267, "isthing": 0, "name": "outlet"},
+    {"color": [255, 9, 92], "id": 268, "isthing": 0, "name": "oven"},
+    {"color": [112, 9, 255], "id": 269, "isthing": 0, "name": "oxygen bottle"},
+    {"color": [8, 255, 214], "id": 270, "isthing": 0, "name": "pack"},
+    {"color": [7, 255, 224], "id": 271, "isthing": 0, "name": "pan"},
+    {"color": [255, 184, 6], "id": 272, "isthing": 0, "name": "paper"},
+    {"color": [10, 255, 71], "id": 273, "isthing": 0, "name": "paper box"},
+    {"color": [255, 41, 10], "id": 274, "isthing": 0, "name": "paper cutter"},
+    {"color": [7, 255, 255], "id": 275, "isthing": 0, "name": "parachute"},
+    {"color": [224, 255, 8], "id": 276, "isthing": 0, "name": "parasol"},
+    {"color": [102, 8, 255], "id": 277, "isthing": 0, "name": "parterre"},
+    {"color": [255, 61, 6], "id": 278, "isthing": 0, "name": "patio"},
+    {"color": [255, 194, 7], "id": 279, "isthing": 0, "name": "pelage"},
+    {"color": [255, 122, 8], "id": 280, "isthing": 0, "name": "pen"},
+    {"color": [0, 255, 20], "id": 281, "isthing": 0, "name": "pen container"},
+    {"color": [255, 8, 41], "id": 282, "isthing": 0, "name": "pencil"},
+    {"color": [255, 5, 153], "id": 283, "isthing": 0, "name": "person"},
+    {"color": [6, 51, 255], "id": 284, "isthing": 0, "name": "photo"},
+    {"color": [235, 12, 255], "id": 285, "isthing": 0, "name": "piano"},
+    {"color": [160, 150, 20], "id": 286, "isthing": 0, "name": "picture"},
+    {"color": [0, 163, 255], "id": 287, "isthing": 0, "name": "pig"},
+    {"color": [140, 140, 140], "id": 288, "isthing": 0, "name": "pillar"},
+    {"color": [250, 10, 15], "id": 289, "isthing": 0, "name": "pillow"},
+    {"color": [20, 255, 0], "id": 290, "isthing": 0, "name": "pipe"},
+    {"color": [31, 255, 0], "id": 291, "isthing": 0, "name": "pitcher"},
+    {"color": [255, 31, 0], "id": 292, "isthing": 0, "name": "plant"},
+    {"color": [255, 224, 0], "id": 293, "isthing": 0, "name": "plastic"},
+    {"color": [153, 255, 0], "id": 294, "isthing": 0, "name": "plate"},
+    {"color": [0, 0, 255], "id": 295, "isthing": 0, "name": "platform"},
+    {"color": [255, 71, 0], "id": 296, "isthing": 0, "name": "player"},
+    {"color": [0, 235, 255], "id": 297, "isthing": 0, "name": "playground"},
+    {"color": [0, 173, 255], "id": 298, "isthing": 0, "name": "pliers"},
+    {"color": [31, 0, 255], "id": 299, "isthing": 0, "name": "plume"},
+    {"color": [120, 120, 120], "id": 300, "isthing": 0, "name": "poker"},
+    {"color": [180, 120, 120], "id": 301, "isthing": 0, "name": "poker chip"},
+    {"color": [6, 230, 230], "id": 302, "isthing": 0, "name": "pole"},
+    {"color": [80, 50, 50], "id": 303, "isthing": 0, "name": "pool table"},
+    {"color": [4, 200, 3], "id": 304, "isthing": 0, "name": "postcard"},
+    {"color": [120, 120, 80], "id": 305, "isthing": 0, "name": "poster"},
+    {"color": [140, 140, 140], "id": 306, "isthing": 0, "name": "pot"},
+    {"color": [204, 5, 255], "id": 307, "isthing": 0, "name": "pottedplant"},
+    {"color": [230, 230, 230], "id": 308, "isthing": 0, "name": "printer"},
+    {"color": [4, 250, 7], "id": 309, "isthing": 0, "name": "projector"},
+    {"color": [224, 5, 255], "id": 310, "isthing": 0, "name": "pumpkin"},
+    {"color": [235, 255, 7], "id": 311, "isthing": 0, "name": "rabbit"},
+    {"color": [150, 5, 61], "id": 312, "isthing": 0, "name": "racket"},
+    {"color": [120, 120, 70], "id": 313, "isthing": 0, "name": "radiator"},
+    {"color": [8, 255, 51], "id": 314, "isthing": 0, "name": "radio"},
+    {"color": [255, 6, 82], "id": 315, "isthing": 0, "name": "rail"},
+    {"color": [143, 255, 140], "id": 316, "isthing": 0, "name": "rake"},
+    {"color": [204, 255, 4], "id": 317, "isthing": 0, "name": "ramp"},
+    {"color": [255, 51, 7], "id": 318, "isthing": 0, "name": "range hood"},
+    {"color": [204, 70, 3], "id": 319, "isthing": 0, "name": "receiver"},
+    {"color": [0, 102, 200], "id": 320, "isthing": 0, "name": "recorder"},
+    {"color": [61, 230, 250], "id": 321, "isthing": 0, "name": "recreational machines"},
+    {"color": [255, 6, 51], "id": 322, "isthing": 0, "name": "remote control"},
+    {"color": [11, 102, 255], "id": 323, "isthing": 0, "name": "road"},
+    {"color": [255, 7, 71], "id": 324, "isthing": 0, "name": "robot"},
+    {"color": [255, 9, 224], "id": 325, "isthing": 0, "name": "rock"},
+    {"color": [9, 7, 230], "id": 326, "isthing": 0, "name": "rocket"},
+    {"color": [220, 220, 220], "id": 327, "isthing": 0, "name": "rocking horse"},
+    {"color": [255, 9, 92], "id": 328, "isthing": 0, "name": "rope"},
+    {"color": [112, 9, 255], "id": 329, "isthing": 0, "name": "rug"},
+    {"color": [8, 255, 214], "id": 330, "isthing": 0, "name": "ruler"},
+    {"color": [7, 255, 224], "id": 331, "isthing": 0, "name": "runway"},
+    {"color": [255, 184, 6], "id": 332, "isthing": 0, "name": "saddle"},
+    {"color": [10, 255, 71], "id": 333, "isthing": 0, "name": "sand"},
+    {"color": [255, 41, 10], "id": 334, "isthing": 0, "name": "saw"},
+    {"color": [7, 255, 255], "id": 335, "isthing": 0, "name": "scale"},
+    {"color": [224, 255, 8], "id": 336, "isthing": 0, "name": "scanner"},
+    {"color": [102, 8, 255], "id": 337, "isthing": 0, "name": "scissors"},
+    {"color": [255, 61, 6], "id": 338, "isthing": 0, "name": "scoop"},
+    {"color": [255, 194, 7], "id": 339, "isthing": 0, "name": "screen"},
+    {"color": [255, 122, 8], "id": 340, "isthing": 0, "name": "screwdriver"},
+    {"color": [0, 255, 20], "id": 341, "isthing": 0, "name": "sculpture"},
+    {"color": [255, 8, 41], "id": 342, "isthing": 0, "name": "scythe"},
+    {"color": [255, 5, 153], "id": 343, "isthing": 0, "name": "sewer"},
+    {"color": [6, 51, 255], "id": 344, "isthing": 0, "name": "sewing machine"},
+    {"color": [235, 12, 255], "id": 345, "isthing": 0, "name": "shed"},
+    {"color": [160, 150, 20], "id": 346, "isthing": 0, "name": "sheep"},
+    {"color": [0, 163, 255], "id": 347, "isthing": 0, "name": "shell"},
+    {"color": [140, 140, 140], "id": 348, "isthing": 0, "name": "shelves"},
+    {"color": [250, 10, 15], "id": 349, "isthing": 0, "name": "shoe"},
+    {"color": [20, 255, 0], "id": 350, "isthing": 0, "name": "shopping cart"},
+    {"color": [31, 255, 0], "id": 351, "isthing": 0, "name": "shovel"},
+    {"color": [255, 31, 0], "id": 352, "isthing": 0, "name": "sidecar"},
+    {"color": [255, 224, 0], "id": 353, "isthing": 0, "name": "sidewalk"},
+    {"color": [153, 255, 0], "id": 354, "isthing": 0, "name": "sign"},
+    {"color": [0, 0, 255], "id": 355, "isthing": 0, "name": "signal light"},
+    {"color": [255, 71, 0], "id": 356, "isthing": 0, "name": "sink"},
+    {"color": [0, 235, 255], "id": 357, "isthing": 0, "name": "skateboard"},
+    {"color": [0, 173, 255], "id": 358, "isthing": 0, "name": "ski"},
+    {"color": [31, 0, 255], "id": 359, "isthing": 0, "name": "sky"},
+    {"color": [120, 120, 120], "id": 360, "isthing": 0, "name": "sled"},
+    {"color": [180, 120, 120], "id": 361, "isthing": 0, "name": "slippers"},
+    {"color": [6, 230, 230], "id": 362, "isthing": 0, "name": "smoke"},
+    {"color": [80, 50, 50], "id": 363, "isthing": 0, "name": "snail"},
+    {"color": [4, 200, 3], "id": 364, "isthing": 0, "name": "snake"},
+    {"color": [120, 120, 80], "id": 365, "isthing": 0, "name": "snow"},
+    {"color": [140, 140, 140], "id": 366, "isthing": 0, "name": "snowmobiles"},
+    {"color": [204, 5, 255], "id": 367, "isthing": 0, "name": "sofa"},
+    {"color": [230, 230, 230], "id": 368, "isthing": 0, "name": "spanner"},
+    {"color": [4, 250, 7], "id": 369, "isthing": 0, "name": "spatula"},
+    {"color": [224, 5, 255], "id": 370, "isthing": 0, "name": "speaker"},
+    {"color": [235, 255, 7], "id": 371, "isthing": 0, "name": "speed bump"},
+    {"color": [150, 5, 61], "id": 372, "isthing": 0, "name": "spice container"},
+    {"color": [120, 120, 70], "id": 373, "isthing": 0, "name": "spoon"},
+    {"color": [8, 255, 51], "id": 374, "isthing": 0, "name": "sprayer"},
+    {"color": [255, 6, 82], "id": 375, "isthing": 0, "name": "squirrel"},
+    {"color": [143, 255, 140], "id": 376, "isthing": 0, "name": "stage"},
+    {"color": [204, 255, 4], "id": 377, "isthing": 0, "name": "stair"},
+    {"color": [255, 51, 7], "id": 378, "isthing": 0, "name": "stapler"},
+    {"color": [204, 70, 3], "id": 379, "isthing": 0, "name": "stick"},
+    {"color": [0, 102, 200], "id": 380, "isthing": 0, "name": "sticky note"},
+    {"color": [61, 230, 250], "id": 381, "isthing": 0, "name": "stone"},
+    {"color": [255, 6, 51], "id": 382, "isthing": 0, "name": "stool"},
+    {"color": [11, 102, 255], "id": 383, "isthing": 0, "name": "stove"},
+    {"color": [255, 7, 71], "id": 384, "isthing": 0, "name": "straw"},
+    {"color": [255, 9, 224], "id": 385, "isthing": 0, "name": "stretcher"},
+    {"color": [9, 7, 230], "id": 386, "isthing": 0, "name": "sun"},
+    {"color": [220, 220, 220], "id": 387, "isthing": 0, "name": "sunglass"},
+    {"color": [255, 9, 92], "id": 388, "isthing": 0, "name": "sunshade"},
+    {"color": [112, 9, 255], "id": 389, "isthing": 0, "name": "surveillance camera"},
+    {"color": [8, 255, 214], "id": 390, "isthing": 0, "name": "swan"},
+    {"color": [7, 255, 224], "id": 391, "isthing": 0, "name": "sweeper"},
+    {"color": [255, 184, 6], "id": 392, "isthing": 0, "name": "swim ring"},
+    {"color": [10, 255, 71], "id": 393, "isthing": 0, "name": "swimming pool"},
+    {"color": [255, 41, 10], "id": 394, "isthing": 0, "name": "swing"},
+    {"color": [7, 255, 255], "id": 395, "isthing": 0, "name": "switch"},
+    {"color": [224, 255, 8], "id": 396, "isthing": 0, "name": "table"},
+    {"color": [102, 8, 255], "id": 397, "isthing": 0, "name": "tableware"},
+    {"color": [255, 61, 6], "id": 398, "isthing": 0, "name": "tank"},
+    {"color": [255, 194, 7], "id": 399, "isthing": 0, "name": "tap"},
+    {"color": [255, 122, 8], "id": 400, "isthing": 0, "name": "tape"},
+    {"color": [0, 255, 20], "id": 401, "isthing": 0, "name": "tarp"},
+    {"color": [255, 8, 41], "id": 402, "isthing": 0, "name": "telephone"},
+    {"color": [255, 5, 153], "id": 403, "isthing": 0, "name": "telephone booth"},
+    {"color": [6, 51, 255], "id": 404, "isthing": 0, "name": "tent"},
+    {"color": [235, 12, 255], "id": 405, "isthing": 0, "name": "tire"},
+    {"color": [160, 150, 20], "id": 406, "isthing": 0, "name": "toaster"},
+    {"color": [0, 163, 255], "id": 407, "isthing": 0, "name": "toilet"},
+    {"color": [140, 140, 140], "id": 408, "isthing": 0, "name": "tong"},
+    {"color": [250, 10, 15], "id": 409, "isthing": 0, "name": "tool"},
+    {"color": [20, 255, 0], "id": 410, "isthing": 0, "name": "toothbrush"},
+    {"color": [31, 255, 0], "id": 411, "isthing": 0, "name": "towel"},
+    {"color": [255, 31, 0], "id": 412, "isthing": 0, "name": "toy"},
+    {"color": [255, 224, 0], "id": 413, "isthing": 0, "name": "toy car"},
+    {"color": [153, 255, 0], "id": 414, "isthing": 0, "name": "track"},
+    {"color": [0, 0, 255], "id": 415, "isthing": 0, "name": "train"},
+    {"color": [255, 71, 0], "id": 416, "isthing": 0, "name": "trampoline"},
+    {"color": [0, 235, 255], "id": 417, "isthing": 0, "name": "trash bin"},
+    {"color": [0, 173, 255], "id": 418, "isthing": 0, "name": "tray"},
+    {"color": [31, 0, 255], "id": 419, "isthing": 0, "name": "tree"},
+    {"color": [120, 120, 120], "id": 420, "isthing": 0, "name": "tricycle"},
+    {"color": [180, 120, 120], "id": 421, "isthing": 0, "name": "tripod"},
+    {"color": [6, 230, 230], "id": 422, "isthing": 0, "name": "trophy"},
+    {"color": [80, 50, 50], "id": 423, "isthing": 0, "name": "truck"},
+    {"color": [4, 200, 3], "id": 424, "isthing": 0, "name": "tube"},
+    {"color": [120, 120, 80], "id": 425, "isthing": 0, "name": "turtle"},
+    {"color": [140, 140, 140], "id": 426, "isthing": 0, "name": "tvmonitor"},
+    {"color": [204, 5, 255], "id": 427, "isthing": 0, "name": "tweezers"},
+    {"color": [230, 230, 230], "id": 428, "isthing": 0, "name": "typewriter"},
+    {"color": [4, 250, 7], "id": 429, "isthing": 0, "name": "umbrella"},
+    {"color": [224, 5, 255], "id": 430, "isthing": 0, "name": "unknown"},
+    {"color": [235, 255, 7], "id": 431, "isthing": 0, "name": "vacuum cleaner"},
+    {"color": [150, 5, 61], "id": 432, "isthing": 0, "name": "vending machine"},
+    {"color": [120, 120, 70], "id": 433, "isthing": 0, "name": "video camera"},
+    {"color": [8, 255, 51], "id": 434, "isthing": 0, "name": "video game console"},
+    {"color": [255, 6, 82], "id": 435, "isthing": 0, "name": "video player"},
+    {"color": [143, 255, 140], "id": 436, "isthing": 0, "name": "video tape"},
+    {"color": [204, 255, 4], "id": 437, "isthing": 0, "name": "violin"},
+    {"color": [255, 51, 7], "id": 438, "isthing": 0, "name": "wakeboard"},
+    {"color": [204, 70, 3], "id": 439, "isthing": 0, "name": "wall"},
+    {"color": [0, 102, 200], "id": 440, "isthing": 0, "name": "wallet"},
+    {"color": [61, 230, 250], "id": 441, "isthing": 0, "name": "wardrobe"},
+    {"color": [255, 6, 51], "id": 442, "isthing": 0, "name": "washing machine"},
+    {"color": [11, 102, 255], "id": 443, "isthing": 0, "name": "watch"},
+    {"color": [255, 7, 71], "id": 444, "isthing": 0, "name": "water"},
+    {"color": [255, 9, 224], "id": 445, "isthing": 0, "name": "water dispenser"},
+    {"color": [9, 7, 230], "id": 446, "isthing": 0, "name": "water pipe"},
+    {"color": [220, 220, 220], "id": 447, "isthing": 0, "name": "water skate board"},
+    {"color": [255, 9, 92], "id": 448, "isthing": 0, "name": "watermelon"},
+    {"color": [112, 9, 255], "id": 449, "isthing": 0, "name": "whale"},
+    {"color": [8, 255, 214], "id": 450, "isthing": 0, "name": "wharf"},
+    {"color": [7, 255, 224], "id": 451, "isthing": 0, "name": "wheel"},
+    {"color": [255, 184, 6], "id": 452, "isthing": 0, "name": "wheelchair"},
+    {"color": [10, 255, 71], "id": 453, "isthing": 0, "name": "window"},
+    {"color": [255, 41, 10], "id": 454, "isthing": 0, "name": "window blinds"},
+    {"color": [7, 255, 255], "id": 455, "isthing": 0, "name": "wineglass"},
+    {"color": [224, 255, 8], "id": 456, "isthing": 0, "name": "wire"},
+    {"color": [102, 8, 255], "id": 457, "isthing": 0, "name": "wood"},
+    {"color": [255, 61, 6], "id": 458, "isthing": 0, "name": "wool"},
+]
+
+PASCAL_VOC_21_CATEGORIES = [
+    {"color": [0, 0, 0], "id": 0, "isthing": 1, "name": "background"},
+    {"color": [128, 0, 0], "id": 1, "isthing": 1, "name": "aeroplane"},
+    {"color": [0, 128, 0], "id": 2, "isthing": 1, "name": "bicycle"},
+    {"color": [128, 128, 0], "id": 3, "isthing": 1, "name": "bird"},
+    {"color": [0, 0, 128], "id": 4, "isthing": 1, "name": "boat"},
+    {"color": [128, 0, 128], "id": 5, "isthing": 1, "name": "bottle"},
+    {"color": [0, 128, 128], "id": 6, "isthing": 1, "name": "bus"},
+    {"color": [128, 128, 128], "id": 7, "isthing": 1, "name": "car"},
+    {"color": [64, 0, 0], "id": 8, "isthing": 1, "name": "cat"},
+    {"color": [192, 0, 0], "id": 9, "isthing": 1, "name": "chair"},
+    {"color": [64, 128, 0], "id": 10, "isthing": 1, "name": "cow"},
+    {"color": [192, 128, 0], "id": 11, "isthing": 1, "name": "diningtable"},
+    {"color": [64, 0, 128], "id": 12, "isthing": 1, "name": "dog"},
+    {"color": [192, 0, 128], "id": 13, "isthing": 1, "name": "horse"},
+    {"color": [64, 128, 128], "id": 14, "isthing": 1, "name": "motorbike"},
+    {"color": [192, 128, 128], "id": 15, "isthing": 1, "name": "person"},
+    {"color": [0, 64, 0], "id": 16, "isthing": 1, "name": "pottedplant"},
+    {"color": [128, 64, 0], "id": 17, "isthing": 1, "name": "sheep"},
+    {"color": [0, 192, 0], "id": 18, "isthing": 1, "name": "sofa"},
+    {"color": [128, 192, 0], "id": 19, "isthing": 1, "name": "train"},
+    {"color": [0, 64, 128], "id": 20, "isthing": 1, "name": "tvmonitor"},
+]
+
+PASCAL_PARTS_CATEGORIES = [
+    {"id": 1, "name": "aeroplane body", "color": [231, 4, 237]},
+    {"id": 2, "name": "aeroplane stern", "color": [116, 80, 69]},
+    {"id": 3, "name": "aeroplane wing", "color": [214, 86, 123]},
+    {"id": 4, "name": "aeroplane tail", "color": [22, 174, 172]},
+    {"id": 5, "name": "aeroplane engine", "color": [197, 128, 182]},
+    {"id": 6, "name": "aeroplane wheel", "color": [82, 197, 247]},
+    {"id": 7, "name": "bicycle body", "color": [125, 34, 155]},
+    {"id": 8, "name": "bicycle wheel", "color": [240, 6, 206]},
+    {"id": 9, "name": "bicycle saddle", "color": [0, 67, 113]},
+    {"id": 10, "name": "bicycle handlebar", "color": [112, 158, 137]},
+    {"id": 11, "name": "bicycle headlight", "color": [255, 182, 87]},
+    {"id": 12, "name": "bird torso", "color": [189, 249, 133]},
+    {"id": 13, "name": "bird head", "color": [104, 202, 100]},
+    {"id": 14, "name": "bird neck", "color": [158, 181, 70]},
+    {"id": 15, "name": "bird wing", "color": [61, 245, 238]},
+    {"id": 16, "name": "bird leg", "color": [55, 126, 0]},
+    {"id": 17, "name": "bird foot", "color": [225, 182, 182]},
+    {"id": 18, "name": "bird tail", "color": [68, 62, 33]},
+    {"id": 19, "name": "boat ", "color": [200, 219, 162]},
+    {"id": 20, "name": "bottle body", "color": [172, 155, 96]},
+    {"id": 21, "name": "bottle cap", "color": [185, 14, 216]},
+    {"id": 22, "name": "bus body", "color": [3, 58, 66]},
+    {"id": 23, "name": "bus frontside", "color": [26, 173, 31]},
+    {"id": 24, "name": "bus leftside", "color": [205, 197, 47]},
+    {"id": 25, "name": "bus rightside", "color": [6, 223, 194]},
+    {"id": 26, "name": "bus backside", "color": [10, 232, 224]},
+    {"id": 27, "name": "bus roofside", "color": [189, 124, 163]},
+    {"id": 28, "name": "bus mirror", "color": [253, 98, 118]},
+    {"id": 29, "name": "bus fliplate", "color": [134, 124, 251]},
+    {"id": 30, "name": "bus bliplate", "color": [86, 248, 252]},
+    {"id": 31, "name": "bus door", "color": [104, 232, 186]},
+    {"id": 32, "name": "bus wheel", "color": [73, 10, 81]},
+    {"id": 33, "name": "bus headlight", "color": [83, 15, 206]},
+    {"id": 34, "name": "bus window", "color": [182, 248, 35]},
+    {"id": 35, "name": "car body", "color": [111, 175, 136]},
+    {"id": 36, "name": "car mirror", "color": [244, 27, 39]},
+    {"id": 37, "name": "car tmirror", "color": [60, 75, 197]},
+    {"id": 38, "name": "car fliplate", "color": [32, 124, 177]},
+    {"id": 39, "name": "car bliplate", "color": [132, 107, 137]},
+    {"id": 40, "name": "car door", "color": [29, 145, 220]},
+    {"id": 41, "name": "car wheel", "color": [211, 58, 216]},
+    {"id": 42, "name": "car headlight", "color": [253, 195, 114]},
+    {"id": 43, "name": "car window", "color": [51, 163, 166]},
+    {"id": 44, "name": "cat torso", "color": [68, 44, 17]},
+    {"id": 45, "name": "cat head", "color": [148, 109, 203]},
+    {"id": 46, "name": "cat eye", "color": [221, 235, 212]},
+    {"id": 47, "name": "cat ear", "color": [25, 226, 114]},
+    {"id": 48, "name": "cat nose", "color": [99, 126, 184]},
+    {"id": 49, "name": "cat neck", "color": [54, 164, 161]},
+    {"id": 50, "name": "cat leg", "color": [114, 251, 219]},
+    {"id": 51, "name": "cat pawn", "color": [145, 28, 176]},
+    {"id": 52, "name": "cat tail", "color": [22, 29, 245]},
+    {"id": 53, "name": "chair ", "color": [174, 108, 109]},
+    {"id": 54, "name": "cow torso", "color": [153, 207, 125]},
+    {"id": 55, "name": "cow head", "color": [243, 197, 251]},
+    {"id": 56, "name": "cow eye", "color": [99, 87, 120]},
+    {"id": 57, "name": "cow ear", "color": [194, 7, 114]},
+    {"id": 58, "name": "cow muzzle", "color": [242, 122, 177]},
+    {"id": 59, "name": "cow horn", "color": [202, 242, 232]},
+    {"id": 60, "name": "cow neck", "color": [250, 136, 178]},
+    {"id": 61, "name": "cow leg", "color": [171, 46, 206]},
+    {"id": 62, "name": "cow tail", "color": [186, 133, 2]},
+    {"id": 63, "name": "table ", "color": [201, 1, 108]},
+    {"id": 64, "name": "dog torso", "color": [245, 11, 186]},
+    {"id": 65, "name": "dog head", "color": [33, 191, 131]},
+    {"id": 66, "name": "dog eye", "color": [225, 95, 66]},
+    {"id": 67, "name": "dog ear", "color": [124, 25, 24]},
+    {"id": 68, "name": "dog nose", "color": [214, 234, 112]},
+    {"id": 69, "name": "dog neck", "color": [129, 83, 21]},
+    {"id": 70, "name": "dog leg", "color": [185, 76, 143]},
+    {"id": 71, "name": "dog pawn", "color": [180, 1, 74]},
+    {"id": 72, "name": "dog tail", "color": [121, 134, 63]},
+    {"id": 73, "name": "dog muzzle", "color": [90, 58, 214]},
+    {"id": 74, "name": "horse body", "color": [223, 7, 152]},
+    {"id": 75, "name": "horse head", "color": [154, 96, 130]},
+    {"id": 76, "name": "horse eye", "color": [221, 98, 183]},
+    {"id": 77, "name": "horse ear", "color": [230, 145, 183]},
+    {"id": 78, "name": "horse muzzle", "color": [213, 203, 88]},
+    {"id": 79, "name": "horse torso", "color": [183, 92, 254]},
+    {"id": 80, "name": "horse neck", "color": [206, 114, 11]},
+    {"id": 81, "name": "horse leg", "color": [214, 238, 15]},
+    {"id": 82, "name": "horse tail", "color": [57, 239, 109]},
+    {"id": 83, "name": "motorbike body", "color": [197, 138, 146]},
+    {"id": 84, "name": "motorbike wheel", "color": [124, 107, 252]},
+    {"id": 85, "name": "motorbike handlebar", "color": [163, 225, 169]},
+    {"id": 86, "name": "motorbike saddle", "color": [254, 180, 116]},
+    {"id": 87, "name": "motorbike headlight", "color": [119, 52, 22]},
+    {"id": 88, "name": "person body", "color": [198, 68, 18]},
+    {"id": 89, "name": "person head", "color": [40, 30, 77]},
+    {"id": 90, "name": "person eye", "color": [237, 64, 148]},
+    {"id": 91, "name": "person ear", "color": [49, 186, 234]},
+    {"id": 92, "name": "person ebrow", "color": [242, 204, 127]},
+    {"id": 93, "name": "person nose", "color": [101, 145, 176]},
+    {"id": 94, "name": "person mouth", "color": [31, 78, 216]},
+    {"id": 95, "name": "person hair", "color": [95, 148, 151]},
+    {"id": 96, "name": "person torso", "color": [126, 117, 235]},
+    {"id": 97, "name": "person neck", "color": [13, 146, 62]},
+    {"id": 98, "name": "person lower arm", "color": [9, 41, 5]},
+    {"id": 99, "name": "person upper arm", "color": [110, 109, 109]},
+    {"id": 100, "name": "person hand", "color": [58, 227, 163]},
+    {"id": 101, "name": "person lower leg", "color": [132, 63, 32]},
+    {"id": 102, "name": "person upper leg", "color": [212, 118, 174]},
+    {"id": 103, "name": "person foot", "color": [45, 66, 254]},
+    {"id": 104, "name": "pottedplant plant", "color": [236, 149, 209]},
+    {"id": 105, "name": "pottedplant pot", "color": [80, 197, 134]},
+    {"id": 106, "name": "sheep torso", "color": [241, 111, 194]},
+    {"id": 107, "name": "sheep head", "color": [31, 13, 13]},
+    {"id": 108, "name": "sheep eye", "color": [34, 207, 63]},
+    {"id": 109, "name": "sheep ear", "color": [249, 117, 121]},
+    {"id": 110, "name": "sheep muzzle", "color": [172, 128, 70]},
+    {"id": 111, "name": "sheep horn", "color": [97, 144, 104]},
+    {"id": 112, "name": "sheep neck", "color": [121, 163, 14]},
+    {"id": 113, "name": "sheep leg", "color": [38, 79, 231]},
+    {"id": 114, "name": "sheep tail", "color": [218, 195, 52]},
+    {"id": 115, "name": "sofa ", "color": [102, 8, 225]},
+    {"id": 116, "name": "train body", "color": [150, 44, 180]},
+    {"id": 117, "name": "train head", "color": [99, 250, 180]},
+    {"id": 118, "name": "train headlight", "color": [24, 148, 249]},
+    {"id": 119, "name": "train coach", "color": [143, 232, 181]},
+    {"id": 120, "name": "tvmonitor frame", "color": [68, 191, 134]},
+    {"id": 121, "name": "tvmonitor screen", "color": [186, 6, 38]},
+    {"id": 122, "name": "bag ", "color": [215, 253, 9]},
+    {"id": 123, "name": "basket ", "color": [150, 44, 154]},
+    {"id": 124, "name": "bed ", "color": [66, 132, 108]},
+    {"id": 125, "name": "bedclothes ", "color": [193, 84, 92]},
+    {"id": 126, "name": "bench ", "color": [84, 154, 254]},
+    {"id": 127, "name": "bird cage ", "color": [2, 93, 169]},
+    {"id": 128, "name": "board ", "color": [41, 254, 95]},
+    {"id": 129, "name": "book ", "color": [157, 228, 148]},
+    {"id": 130, "name": "bowl ", "color": [201, 198, 2]},
+    {"id": 131, "name": "box ", "color": [237, 151, 223]},
+    {"id": 132, "name": "bridge ", "color": [74, 200, 197]},
+    {"id": 133, "name": "brush ", "color": [157, 2, 192]},
+    {"id": 134, "name": "bucket ", "color": [62, 8, 145]},
+    {"id": 135, "name": "building ", "color": [244, 158, 23]},
+    {"id": 136, "name": "cabinet ", "color": [143, 34, 160]},
+    {"id": 137, "name": "cage ", "color": [74, 182, 153]},
+    {"id": 138, "name": "case ", "color": [44, 161, 32]},
+    {"id": 139, "name": "ceiling ", "color": [22, 207, 172]},
+    {"id": 140, "name": "cloth ", "color": [62, 233, 51]},
+    {"id": 141, "name": "computer ", "color": [203, 221, 8]},
+    {"id": 142, "name": "counter ", "color": [155, 154, 208]},
+    {"id": 143, "name": "cup ", "color": [136, 170, 161]},
+    {"id": 144, "name": "curtain ", "color": [69, 238, 67]},
+    {"id": 145, "name": "cushion ", "color": [250, 140, 63]},
+    {"id": 146, "name": "door ", "color": [228, 29, 142]},
+    {"id": 147, "name": "fence ", "color": [149, 149, 255]},
+    {"id": 148, "name": "fire place ", "color": [25, 17, 14]},
+    {"id": 149, "name": "floor ", "color": [141, 121, 107]},
+    {"id": 150, "name": "flower ", "color": [196, 171, 99]},
+    {"id": 151, "name": "food ", "color": [246, 30, 195]},
+    {"id": 152, "name": "fridge ", "color": [95, 2, 42]},
+    {"id": 153, "name": "grandstand ", "color": [174, 116, 162]},
+    {"id": 154, "name": "grass ", "color": [251, 58, 246]},
+    {"id": 155, "name": "ground ", "color": [138, 68, 168]},
+    {"id": 156, "name": "horse-drawn carriage ", "color": [236, 220, 194]},
+    {"id": 157, "name": "keyboard ", "color": [228, 180, 129]},
+    {"id": 158, "name": "laptop ", "color": [41, 39, 187]},
+    {"id": 159, "name": "light ", "color": [18, 155, 71]},
+    {"id": 160, "name": "mat ", "color": [81, 149, 168]},
+    {"id": 161, "name": "metal ", "color": [222, 250, 122]},
+    {"id": 162, "name": "mirror ", "color": [27, 14, 162]},
+    {"id": 163, "name": "mountain ", "color": [96, 67, 42]},
+    {"id": 164, "name": "mouse ", "color": [248, 27, 142]},
+    {"id": 165, "name": "pack ", "color": [48, 208, 79]},
+    {"id": 166, "name": "paper ", "color": [85, 44, 114]},
+    {"id": 167, "name": "picture ", "color": [8, 66, 36]},
+    {"id": 168, "name": "pillow ", "color": [199, 38, 36]},
+    {"id": 169, "name": "plant ", "color": [45, 67, 214]},
+    {"id": 170, "name": "plate ", "color": [176, 85, 199]},
+    {"id": 171, "name": "platform ", "color": [118, 46, 134]},
+    {"id": 172, "name": "pole ", "color": [66, 53, 97]},
+    {"id": 173, "name": "poster ", "color": [134, 95, 198]},
+    {"id": 174, "name": "pot ", "color": [56, 185, 27]},
+    {"id": 175, "name": "road ", "color": [12, 12, 242]},
+    {"id": 176, "name": "rock ", "color": [141, 182, 239]},
+    {"id": 177, "name": "rope ", "color": [242, 15, 134]},
+    {"id": 178, "name": "rug ", "color": [119, 78, 116]},
+    {"id": 179, "name": "sand ", "color": [159, 25, 177]},
+    {"id": 180, "name": "sculpture ", "color": [155, 71, 2]},
+    {"id": 181, "name": "shelves ", "color": [13, 156, 172]},
+    {"id": 182, "name": "sidewalk ", "color": [153, 56, 74]},
+    {"id": 183, "name": "sign ", "color": [132, 5, 169]},
+    {"id": 184, "name": "sink ", "color": [202, 115, 244]},
+    {"id": 185, "name": "sky ", "color": [189, 81, 126]},
+    {"id": 186, "name": "smoke ", "color": [50, 105, 141]},
+    {"id": 187, "name": "snow ", "color": [163, 75, 126]},
+    {"id": 188, "name": "speaker ", "color": [25, 28, 9]},
+    {"id": 189, "name": "stage ", "color": [57, 175, 211]},
+    {"id": 190, "name": "stair ", "color": [36, 182, 123]},
+    {"id": 191, "name": "tent ", "color": [210, 184, 159]},
+    {"id": 192, "name": "toy ", "color": [139, 14, 196]},
+    {"id": 193, "name": "track ", "color": [204, 225, 55]},
+    {"id": 194, "name": "tree ", "color": [145, 64, 92]},
+    {"id": 195, "name": "truck ", "color": [43, 65, 241]},
+    {"id": 196, "name": "wall ", "color": [220, 189, 61]},
+    {"id": 197, "name": "water ", "color": [250, 95, 220]},
+    {"id": 198, "name": "window ", "color": [176, 117, 245]},
+    {"id": 199, "name": "wineglass ", "color": [102, 162, 66]},
+    {"id": 200, "name": "wood ", "color": [100, 60, 45]},
+]
+
+PASCAL_PARTS_PARTS_ONLY = [
+    {"id": 1, "name": "aeroplane body", "color": [30, 178, 112]},
+    {"id": 2, "name": "aeroplane stern", "color": [0, 80, 42]},
+    {"id": 3, "name": "aeroplane wing", "color": [160, 237, 245]},
+    {"id": 4, "name": "aeroplane engine", "color": [144, 222, 51]},
+    {"id": 5, "name": "aeroplane wheel", "color": [155, 121, 20]},
+    {"id": 6, "name": "bicycle body", "color": [24, 50, 96]},
+    {"id": 7, "name": "bicycle wheel", "color": [247, 201, 171]},
+    {"id": 8, "name": "bird torso", "color": [63, 162, 50]},
+    {"id": 9, "name": "bird head", "color": [143, 18, 27]},
+    {"id": 10, "name": "bird wing", "color": [204, 34, 128]},
+    {"id": 11, "name": "bird leg", "color": [31, 37, 39]},
+    {"id": 12, "name": "boat ", "color": [175, 16, 226]},
+    {"id": 13, "name": "bottle cap", "color": [31, 13, 221]},
+    {"id": 14, "name": "bottle body", "color": [120, 248, 33]},
+    {"id": 15, "name": "bus body", "color": [243, 104, 244]},
+    {"id": 16, "name": "bus wheel", "color": [247, 196, 104]},
+    {"id": 17, "name": "bus window", "color": [63, 138, 111]},
+    {"id": 18, "name": "car body", "color": [200, 176, 116]},
+    {"id": 19, "name": "car license plate", "color": [79, 146, 205]},
+    {"id": 20, "name": "car wheel", "color": [231, 126, 229]},
+    {"id": 21, "name": "car light", "color": [120, 219, 85]},
+    {"id": 22, "name": "car window", "color": [240, 73, 236]},
+    {"id": 23, "name": "cat torso", "color": [24, 254, 246]},
+    {"id": 24, "name": "cat head", "color": [38, 29, 151]},
+    {"id": 25, "name": "cat leg", "color": [229, 8, 161]},
+    {"id": 26, "name": "cat tail", "color": [212, 191, 142]},
+    {"id": 27, "name": "chair ", "color": [235, 90, 210]},
+    {"id": 28, "name": "cow torso", "color": [72, 26, 132]},
+    {"id": 29, "name": "cow head", "color": [28, 249, 68]},
+    {"id": 30, "name": "cow leg", "color": [69, 62, 39]},
+    {"id": 31, "name": "cow tail", "color": [238, 140, 59]},
+    {"id": 32, "name": "table ", "color": [73, 170, 102]},
+    {"id": 33, "name": "dog torso", "color": [51, 140, 200]},
+    {"id": 34, "name": "dog head", "color": [141, 130, 240]},
+    {"id": 35, "name": "dog leg", "color": [223, 199, 36]},
+    {"id": 36, "name": "dog tail", "color": [40, 192, 182]},
+    {"id": 37, "name": "horse torso", "color": [212, 206, 245]},
+    {"id": 38, "name": "horse head", "color": [59, 63, 103]},
+    {"id": 39, "name": "horse leg", "color": [50, 72, 178]},
+    {"id": 40, "name": "horse tail", "color": [49, 64, 103]},
+    {"id": 41, "name": "motorbike body", "color": [226, 39, 217]},
+    {"id": 42, "name": "motorbike wheel", "color": [11, 110, 195]},
+    {"id": 43, "name": "person torso", "color": [155, 219, 139]},
+    {"id": 44, "name": "person head", "color": [168, 137, 15]},
+    {"id": 45, "name": "person lower arm", "color": [187, 194, 167]},
+    {"id": 46, "name": "person upper arm", "color": [60, 80, 21]},
+    {"id": 47, "name": "person lower leg", "color": [180, 219, 17]},
+    {"id": 48, "name": "person upper leg", "color": [240, 249, 227]},
+    {"id": 49, "name": "pottedplant plant", "color": [191, 176, 151]},
+    {"id": 50, "name": "pottedplant pot", "color": [13, 133, 225]},
+    {"id": 51, "name": "sheep torso", "color": [178, 101, 246]},
+    {"id": 52, "name": "sheep head", "color": [52, 108, 42]},
+    {"id": 53, "name": "sheep leg", "color": [92, 169, 47]},
+    {"id": 54, "name": "sofa ", "color": [45, 45, 192]},
+    {"id": 55, "name": "train body", "color": [168, 7, 178]},
+    {"id": 56, "name": "tvmonitor frame", "color": [59, 89, 2]},
+    {"id": 57, "name": "tvmonitor screen", "color": [51, 85, 167]},
+]
+
+
+def _get_ctx59_meta():
+    # Id 0 is reserved for ignore_label, we change ignore_label for 0
+    # to 255 in our pre-processing, so all ids are shifted by 1.
+    stuff_ids = [k["id"] for k in PASCAL_CTX_59_CATEGORIES]
+    assert len(stuff_ids) == 59, len(stuff_ids)
+
+    # For semantic segmentation, this mapping maps from contiguous stuff id
+    # (in [0, 91], used in models) to ids in the dataset (used for processing results)
+    stuff_dataset_id_to_contiguous_id = {k: i for i, k in enumerate(stuff_ids)}
+    stuff_classes = [k["name"] for k in PASCAL_CTX_59_CATEGORIES]
+
+    ret = {
+        "stuff_dataset_id_to_contiguous_id": stuff_dataset_id_to_contiguous_id,
+        "stuff_classes": stuff_classes,
+    }
+    return ret
+
+
+def register_all_ctx59(root):
+    root = os.path.join(root, "pascal_ctx_d2")
+    meta = _get_ctx59_meta()
+    for name, dirname in [("train", "training"), ("val", "validation")]:
+        image_dir = os.path.join(root, "images", dirname)
+        gt_dir = os.path.join(root, "annotations_ctx59", dirname)
+        name = f"ctx59_sem_seg_{name}"
+        DatasetCatalog.register(
+            name,
+            lambda x=image_dir, y=gt_dir: load_sem_seg(
+                y, x, gt_ext="png", image_ext="jpg", dataset_name="pascal_context_59"
+            ),
+        )
+        MetadataCatalog.get(name).set(
+            stuff_classes=meta["stuff_classes"][:],
+            thing_dataset_id_to_contiguous_id={},  # to make Mask2Former happy
+            stuff_dataset_id_to_contiguous_id=meta["stuff_dataset_id_to_contiguous_id"],
+            image_root=image_dir,
+            sem_seg_root=gt_dir,
+            evaluator_type="sem_seg",
+            ignore_label=255,
+        )
+
+
+def _get_pascal21_meta():
+    # Id 0 is reserved for ignore_label, we change ignore_label for 0
+    # to 255 in our pre-processing, so all ids are shifted by 1.
+    stuff_ids = [k["id"] for k in PASCAL_VOC_21_CATEGORIES]
+    assert len(stuff_ids) == 21, len(stuff_ids)
+
+    # For semantic segmentation, this mapping maps from contiguous stuff id
+    # (in [0, 91], used in models) to ids in the dataset (used for processing results)
+    stuff_dataset_id_to_contiguous_id = {k: i for i, k in enumerate(stuff_ids)}
+    stuff_classes = [k["name"] for k in PASCAL_VOC_21_CATEGORIES]
+
+    ret = {
+        "stuff_dataset_id_to_contiguous_id": stuff_dataset_id_to_contiguous_id,
+        "stuff_classes": stuff_classes,
+    }
+    return ret
+
+
+def register_all_pascal21(root):
+    root = os.path.join(root, "pascal_voc_d2")
+    meta = _get_pascal21_meta()
+    for name, dirname in [("train", "training"), ("val", "validation")]:
+        image_dir = os.path.join(root, "images", dirname)
+        gt_dir = os.path.join(root, "annotations_pascal21", dirname)
+        name = f"pascal21_sem_seg_{name}"
+        DatasetCatalog.register(
+            name,
+            lambda x=image_dir, y=gt_dir: load_sem_seg(
+                y, x, gt_ext="png", image_ext="jpg", dataset_name="pascal_voc_21"
+            ),
+        )
+        MetadataCatalog.get(name).set(
+            stuff_classes=meta["stuff_classes"][:],
+            thing_dataset_id_to_contiguous_id={},  # to make Mask2Former happy
+            stuff_dataset_id_to_contiguous_id=meta["stuff_dataset_id_to_contiguous_id"],
+            image_root=image_dir,
+            sem_seg_root=gt_dir,
+            evaluator_type="sem_seg",
+            ignore_label=255,
+        )
+
+
+def _get_ctx459_meta():
+    # Id 0 is reserved for ignore_label, we change ignore_label for 0
+    # to 255 in our pre-processing, so all ids are shifted by 1.
+    stuff_ids = [k["id"] for k in PASCAL_CTX_459_CATEGORIES]
+    assert len(stuff_ids) == 459, len(stuff_ids)
+
+    # For semantic segmentation, this mapping maps from contiguous stuff id
+    # (in [0, 91], used in models) to ids in the dataset (used for processing results)
+    stuff_dataset_id_to_contiguous_id = {k: i for i, k in enumerate(stuff_ids)}
+    stuff_classes = [k["name"] for k in PASCAL_CTX_459_CATEGORIES]
+
+    ret = {
+        "stuff_dataset_id_to_contiguous_id": stuff_dataset_id_to_contiguous_id,
+        "stuff_classes": stuff_classes,
+    }
+    return ret
+
+
+def register_all_ctx459(root):
+    root = os.path.join(root, "pascal_ctx_d2")
+    meta = _get_ctx459_meta()
+    for name, dirname in [("train", "training"), ("val", "validation")]:
+        image_dir = os.path.join(root, "images", dirname)
+        gt_dir = os.path.join(root, "annotations_ctx459", dirname)
+        name = f"ctx459_sem_seg_{name}"
+        DatasetCatalog.register(
+            name,
+            lambda x=image_dir, y=gt_dir: load_sem_seg(
+                y, x, gt_ext="tif", image_ext="jpg", dataset_name="pascal_context_459"
+            ),
+        )
+        MetadataCatalog.get(name).set(
+            stuff_classes=meta["stuff_classes"][:],
+            thing_dataset_id_to_contiguous_id={},  # to make Mask2Former happy
+            stuff_dataset_id_to_contiguous_id=meta["stuff_dataset_id_to_contiguous_id"],
+            image_root=image_dir,
+            sem_seg_root=gt_dir,
+            evaluator_type="sem_seg",
+            ignore_label=65535,  # NOTE: gt is saved in 16-bit TIFF images
+        )
+
+
+def _get_parts_meta():
+    # Id 0 is reserved for ignore_label, we change ignore_label for 0
+    # to 255 in our pre-processing, so all ids are shifted by 1.
+    stuff_ids = [k["id"] for k in PASCAL_PARTS_CATEGORIES]
+    # assert len(stuff_ids) == 459, len(stuff_ids)
+
+    # For semantic segmentation, this mapping maps from contiguous stuff id
+    # (in [0, 91], used in models) to ids in the dataset (used for processing results)
+    stuff_dataset_id_to_contiguous_id = {k: i for i, k in enumerate(stuff_ids)}
+    stuff_classes = [k["name"] for k in PASCAL_PARTS_CATEGORIES]
+
+    ret = {
+        "stuff_dataset_id_to_contiguous_id": stuff_dataset_id_to_contiguous_id,
+        "stuff_classes": stuff_classes,
+    }
+    return ret
+
+
+def _get_parts_only_meta():
+    # Id 0 is reserved for ignore_label, we change ignore_label for 0
+    # to 255 in our pre-processing, so all ids are shifted by 1.
+    stuff_ids = [k["id"] for k in PASCAL_PARTS_PARTS_ONLY]
+    # assert len(stuff_ids) == 459, len(stuff_ids)
+
+    # For semantic segmentation, this mapping maps from contiguous stuff id
+    # (in [0, 91], used in models) to ids in the dataset (used for processing results)
+    stuff_dataset_id_to_contiguous_id = {k: i for i, k in enumerate(stuff_ids)}
+    stuff_classes = [k["name"] for k in PASCAL_PARTS_PARTS_ONLY]
+
+    ret = {
+        "stuff_dataset_id_to_contiguous_id": stuff_dataset_id_to_contiguous_id,
+        "stuff_classes": stuff_classes,
+    }
+    return ret
+
+
+def register_all_pascal_parts_only(root):
+    data_root = root
+    root = os.path.join(root, "pascal_parts")
+    meta = _get_parts_only_meta()
+    for name, dirname in [
+        ("train", "training_merged"),
+        ("val", "validation_merged"),
+        ("test", "test_merged"),
+    ]:
+        image_dir = os.path.join(data_root, "VOCdevkit/VOC2010/JPEGImages")
+        gt_dir = os.path.join(root, "labels", dirname)
+        name = f"pascal_parts_merged_{name}"
+        DatasetCatalog.register(
+            name,
+            lambda x=image_dir, y=gt_dir: load_sem_seg(
+                y, x, gt_ext="tif", image_ext="jpg", dataset_name="pascal_parts_merged"
+            ),
+        )
+        MetadataCatalog.get(name).set(
+            stuff_classes=meta["stuff_classes"][:],
+            thing_dataset_id_to_contiguous_id={},  # to make Mask2Former happy
+            stuff_dataset_id_to_contiguous_id=meta["stuff_dataset_id_to_contiguous_id"],
+            image_root=image_dir,
+            sem_seg_root=gt_dir,
+            evaluator_type="sem_seg",
+            ignore_label=0,  # NOTE: gt is saved in 16-bit TIFF images
+        )
+
+
+PASCAL_LABEL_PART_GROUP = {
+    1: 1,
+    2: 2,
+    3: 3,
+    4: 2,
+    5: 4,
+    6: 5,
+    7: 6,
+    8: 7,
+    9: 6,
+    10: 6,
+    11: 6,
+    12: 8,
+    13: 9,
+    14: 9,
+    15: 10,
+    16: 11,
+    17: 11,
+    18: 8,
+    19: 12,
+    20: 14,
+    21: 13,
+    22: 15,
+    23: 15,
+    24: 15,
+    25: 15,
+    26: 15,
+    27: 15,
+    28: 15,
+    29: 15,
+    30: 15,
+    31: 15,
+    32: 16,
+    33: 15,
+    34: 17,
+    35: 18,
+    36: 18,
+    37: 18,
+    38: 19,
+    39: 19,
+    40: 18,
+    41: 20,
+    42: 21,
+    43: 22,
+    44: 23,
+    45: 24,
+    46: 24,
+    47: 24,
+    48: 24,
+    49: 23,
+    50: 25,
+    51: 25,
+    52: 26,
+    53: 27,
+    54: 28,
+    55: 29,
+    56: 29,
+    57: 29,
+    58: 29,
+    59: 29,
+    60: 28,
+    61: 30,
+    62: 31,
+    63: 32,
+    64: 33,
+    65: 34,
+    66: 34,
+    67: 34,
+    68: 34,
+    69: 33,
+    70: 35,
+    71: 35,
+    72: 36,
+    73: 34,
+    74: 37,
+    75: 38,
+    76: 38,
+    77: 38,
+    78: 38,
+    79: 37,
+    80: 37,
+    81: 39,
+    82: 40,
+    83: 41,
+    84: 42,
+    85: 41,
+    86: 41,
+    87: 41,
+    88: 43,
+    89: 44,
+    90: 44,
+    91: 44,
+    92: 44,
+    93: 44,
+    94: 44,
+    95: 44,
+    96: 43,
+    97: 43,
+    98: 45,
+    99: 46,
+    100: 45,
+    101: 47,
+    102: 48,
+    103: 47,
+    104: 49,
+    105: 50,
+    106: 51,
+    107: 52,
+    108: 52,
+    109: 52,
+    110: 52,
+    111: 52,
+    112: 51,
+    113: 53,
+    114: 51,
+    115: 54,
+    116: 55,
+    117: 55,
+    118: 55,
+    119: 55,
+    120: 56,
+    121: 57,
+}
+
+
+def register_all_pascal_parts(root):
+    data_root = root
+    register_all_pascal_parts_only(data_root)
+    root = os.path.join(root, "pascal_parts")
+    meta = _get_parts_meta()
+    for name, dirname in [("train", "training"), ("val", "validation"), ("test", "test_pano")]:
+        image_dir = os.path.join(data_root, "VOCdevkit/VOC2010/JPEGImages")
+        gt_dir = os.path.join(root, "labels", dirname)
+        name = f"pascal_parts_{name}"
+        DatasetCatalog.register(
+            name, lambda x=image_dir, y=gt_dir: load_sem_seg(y, x, gt_ext="tif", image_ext="jpg")
+        )
+        MetadataCatalog.get(name).set(
+            stuff_classes=meta["stuff_classes"][:],
+            thing_dataset_id_to_contiguous_id={},  # to make Mask2Former happy
+            stuff_dataset_id_to_contiguous_id=meta["stuff_dataset_id_to_contiguous_id"],
+            image_root=image_dir,
+            sem_seg_root=gt_dir,
+            evaluator_type="sem_seg",
+            label_group=PASCAL_LABEL_PART_GROUP,
+            # ignore_label=0,  # NOTE: gt is saved in 16-bit TIFF images
+            ignore_label=255,  # NOTE: gt is saved in 16-bit TIFF images
+        )
+
+
+_PREDEFINED_SPLITS_PASCALVOCPART = {}
+_PREDEFINED_SPLITS_PASCALVOCPART["pascalvocpart"] = {
+    "pascalvocpart": (
+        "VOCdevkit/VOC2010/JPEGImages",
+        "pascal_parts/pascalvocpart_training_instance.json",
+    ),
+    "pascalvocpart_train": (
+        "VOCdevkit/VOC2010/JPEGImages",
+        "pascal_parts/pascalvocpart_training_instance.json",
+    ),
+    "pascalvocpart_val": (
+        "VOCdevkit/VOC2010/JPEGImages",
+        "pascal_parts/pascalvocpart_validation_instance.json",
+    ),
+}
+
+
+def _get_builtin_metadata(dataset_name):
+    return _get_pascalvocpart_metadata([])
+
+    raise KeyError("No built-in metadata for dataset {}".format(dataset_name))
+
+
+def _get_pascalvocpart_metadata(categories):
+    if len(categories) == 0:
+        return {}
+    id_to_name = {x["id"]: x["name"] for x in categories}
+    thing_dataset_id_to_contiguous_id = {i: i for i in range(len(categories))}
+    thing_classes = [id_to_name[k] for k in sorted(id_to_name)]
+    return {
+        "thing_dataset_id_to_contiguous_id": thing_dataset_id_to_contiguous_id,
+        "thing_classes": thing_classes,
+    }
+
+
+def register_all_pascalvocpart(root):
+    for dataset_name, splits_per_dataset in _PREDEFINED_SPLITS_PASCALVOCPART.items():
+        for key, (image_root, json_file) in splits_per_dataset.items():
+            custom_register_coco_instances(
+                key,
+                _get_builtin_metadata(dataset_name),
+                os.path.join(root, json_file) if "://" not in json_file else json_file,
+                os.path.join(root, image_root),
+            )
+
+
+# register_all_ctx59(os.getenv("DETECTRON2_DATASETS", "datasets"))
+# register_all_pascal21(os.getenv("DETECTRON2_DATASETS", "datasets"))
+# register_all_ctx459(os.getenv("DETECTRON2_DATASETS", "datasets"))
+
+# True for open source;
+# Internally at fb, we register them elsewhere
+if __name__.endswith(".pascal_voc_external"):
+    # Assume pre-defined datasets live in `./datasets`.
+    _root = os.path.expanduser(os.getenv("DETECTRON2_DATASETS", "datasets"))
+    # register_all_pascal_parts(_root)
+    register_all_pascalvocpart(_root)

approach/ovod/APE/ape/data/datasets/phrasecut.py

@@ -0,0 +1,60 @@
+import logging
+import os
+
+from .coco import custom_register_coco_instances
+
+logger = logging.getLogger(__name__)
+
+
+def _get_builtin_metadata(dataset_name):
+    return _get_phrasecut_metadata([])
+
+    raise KeyError("No built-in metadata for dataset {}".format(dataset_name))
+
+
+def _get_phrasecut_metadata(categories):
+    if len(categories) == 0:
+        return {}
+    id_to_name = {x["id"]: x["name"] for x in categories}
+    thing_dataset_id_to_contiguous_id = {i + 1: i for i in range(len(categories))}
+    thing_classes = [id_to_name[k] for k in sorted(id_to_name)]
+    return {
+        "thing_dataset_id_to_contiguous_id": thing_dataset_id_to_contiguous_id,
+        "thing_classes": thing_classes,
+    }
+
+
+_PREDEFINED_SPLITS_PHRASECUT = {}
+_PREDEFINED_SPLITS_PHRASECUT["phrasecut"] = {
+    "phrasecut": (
+        "phrasecut/images",
+        "phrasecut/phrasecut.json",
+    ),
+    "phrasecut_train": (
+        "phrasecut/images",
+        "phrasecut/phrasecut_train.json",
+    ),
+    "phrasecut_val": (
+        "phrasecut/images",
+        "phrasecut/phrasecut_val.json",
+    ),
+}
+
+
+def register_all_phrasecut(root):
+    for dataset_name, splits_per_dataset in _PREDEFINED_SPLITS_PHRASECUT.items():
+        for key, (image_root, json_file) in splits_per_dataset.items():
+            custom_register_coco_instances(
+                key,
+                _get_builtin_metadata(dataset_name),
+                os.path.join(root, json_file) if "://" not in json_file else json_file,
+                os.path.join(root, image_root),
+            )
+
+
+# True for open source;
+# Internally at fb, we register them elsewhere
+if __name__.endswith(".phrasecut"):
+    # Assume pre-defined datasets live in `./datasets`.
+    _root = os.path.expanduser(os.getenv("DETECTRON2_DATASETS", "datasets"))
+    register_all_phrasecut(_root)

approach/ovod/APE/ape/data/datasets/refcoco.py

@@ -0,0 +1,337 @@
+import contextlib
+import io
+import logging
+import os
+
+import numpy as np
+import pycocotools.mask as mask_util
+from PIL import Image
+
+from detectron2.data import DatasetCatalog, MetadataCatalog
+from detectron2.structures import Boxes, BoxMode, PolygonMasks, RotatedBoxes
+from detectron2.utils.file_io import PathManager
+from fvcore.common.timer import Timer
+from iopath.common.file_io import file_lock
+
+"""
+This file contains functions to parse COCO-format annotations into dicts in "Detectron2 format".
+"""
+
+
+logger = logging.getLogger(__name__)
+
+__all__ = ["load_refcoco_json", "convert_to_coco_json", "register_refcoco"]
+
+REFCOCO_CATEGORIES = [
+    {"color": [220, 20, 60], "isthing": 1, "id": 1, "name": "object"}
+]  # only one class for visual grounding
+
+
+def _get_refcoco_meta():
+    thing_ids = [k["id"] for k in REFCOCO_CATEGORIES if k["isthing"] == 1]
+    thing_colors = [k["color"] for k in REFCOCO_CATEGORIES if k["isthing"] == 1]
+    assert len(thing_ids) == 1, len(thing_ids)
+
+    thing_dataset_id_to_contiguous_id = {k: i for i, k in enumerate(thing_ids)}
+    thing_classes = [k["name"] for k in REFCOCO_CATEGORIES if k["isthing"] == 1]
+    ret = {
+        "thing_dataset_id_to_contiguous_id": thing_dataset_id_to_contiguous_id,
+        "thing_classes": thing_classes,
+        "thing_colors": thing_colors,
+    }
+    return ret
+
+
+def load_refcoco_json(json_file, image_root, dataset_name=None, extra_annotation_keys=None):
+    """
+    Load a json file with COCO's instances annotation format.
+    Currently supports instance detection, instance segmentation,
+    and person keypoints annotations.
+
+    Args:
+        json_file (str): full path to the json file in COCO instances annotation format.
+        image_root (str or path-like): the directory where the images in this json file exists.
+        dataset_name (str or None): the name of the dataset (e.g., coco_2017_train).
+            When provided, this function will also do the following:
+
+            * Put "thing_classes" into the metadata associated with this dataset.
+            * Map the category ids into a contiguous range (needed by standard dataset format),
+              and add "thing_dataset_id_to_contiguous_id" to the metadata associated
+              with this dataset.
+
+            This option should usually be provided, unless users need to load
+            the original json content and apply more processing manually.
+        extra_annotation_keys (list[str]): list of per-annotation keys that should also be
+            loaded into the dataset dict (besides "iscrowd", "bbox", "keypoints",
+            "category_id", "segmentation"). The values for these keys will be returned as-is.
+            For example, the densepose annotations are loaded in this way.
+
+    Returns:
+        list[dict]: a list of dicts in Detectron2 standard dataset dicts format (See
+        `Using Custom Datasets </tutorials/datasets.html>`_ ) when `dataset_name` is not None.
+        If `dataset_name` is None, the returned `category_ids` may be
+        incontiguous and may not conform to the Detectron2 standard format.
+
+    Notes:
+        1. This function does not read the image files.
+           The results do not have the "image" field.
+    """
+    from pycocotools.coco import COCO
+
+    timer = Timer()
+    json_file = PathManager.get_local_path(json_file)
+    with contextlib.redirect_stdout(io.StringIO()):
+        coco_api = COCO(json_file)
+    if timer.seconds() > 1:
+        logger.info("Loading {} takes {:.2f} seconds.".format(json_file, timer.seconds()))
+
+    id_map = None
+    if dataset_name is not None:
+        meta = MetadataCatalog.get(dataset_name)
+        cat_ids = sorted(coco_api.getCatIds())
+        cats = coco_api.loadCats(cat_ids)
+        # The categories in a custom json file may not be sorted.
+        thing_classes = [c["name"] for c in sorted(cats, key=lambda x: x["id"])]
+        meta.thing_classes = thing_classes
+
+        # In COCO, certain category ids are artificially removed,
+        # and by convention they are always ignored.
+        # We deal with COCO's id issue and translate
+        # the category ids to contiguous ids in [0, 80).
+
+        # It works by looking at the "categories" field in the json, therefore
+        # if users' own json also have incontiguous ids, we'll
+        # apply this mapping as well but print a warning.
+        if not (min(cat_ids) == 1 and max(cat_ids) == len(cat_ids)):
+            if "coco" not in dataset_name:
+                logger.warning(
+                    """
+Category ids in annotations are not in [1, #categories]! We'll apply a mapping for you.
+"""
+                )
+        id_map = {v: i for i, v in enumerate(cat_ids)}
+        meta.thing_dataset_id_to_contiguous_id = id_map
+
+        cat_ids = cat_ids + list(range(max(cat_ids) + 1, 100000))
+        id_map = {v: i for i, v in enumerate(cat_ids)}
+
+    # sort indices for reproducible results
+    img_ids = sorted(coco_api.imgs.keys())
+    # imgs is a list of dicts, each looks something like:
+    # {'license': 4,
+    #  'url': 'http://farm6.staticflickr.com/5454/9413846304_881d5e5c3b_z.jpg',
+    #  'file_name': 'COCO_val2014_000000001268.jpg',
+    #  'height': 427,
+    #  'width': 640,
+    #  'date_captured': '2013-11-17 05:57:24',
+    #  'id': 1268}
+    imgs = coco_api.loadImgs(img_ids)
+    # anns is a list[list[dict]], where each dict is an annotation
+    # record for an object. The inner list enumerates the objects in an image
+    # and the outer list enumerates over images. Example of anns[0]:
+    # [{'segmentation': [[192.81,
+    #     247.09,
+    #     ...
+    #     219.03,
+    #     249.06]],
+    #   'area': 1035.749,
+    #   'iscrowd': 0,
+    #   'image_id': 1268,
+    #   'bbox': [192.81, 224.8, 74.73, 33.43],
+    #   'category_id': 16,
+    #   'id': 42986},
+    #  ...]
+    anns = [coco_api.imgToAnns[img_id] for img_id in img_ids]
+    total_num_valid_anns = sum([len(x) for x in anns])
+    total_num_anns = len(coco_api.anns)
+    if total_num_valid_anns < total_num_anns:
+        logger.warning(
+            f"{json_file} contains {total_num_anns} annotations, but only "
+            f"{total_num_valid_anns} of them match to images in the file."
+        )
+
+    if "minival" not in json_file:
+        # The popular valminusminival & minival annotations for COCO2014 contain this bug.
+        # However the ratio of buggy annotations there is tiny and does not affect accuracy.
+        # Therefore we explicitly white-list them.
+        ann_ids = [ann["id"] for anns_per_image in anns for ann in anns_per_image]
+        assert len(set(ann_ids)) == len(ann_ids), "Annotation ids in '{}' are not unique!".format(
+            json_file
+        )
+
+    imgs_anns = list(zip(imgs, anns))
+    logger.info("Loaded {} images in COCO format from {}".format(len(imgs_anns), json_file))
+
+    dataset_dicts = []
+
+    ann_keys = ["iscrowd", "bbox", "keypoints", "category_id"] + (extra_annotation_keys or [])
+
+    num_instances_without_valid_segmentation = 0
+
+    for (img_dict, anno_dict_list) in imgs_anns:
+        record = {}
+        record["file_name"] = os.path.join(image_root, img_dict["file_name"])
+        record["height"] = img_dict["height"]
+        record["width"] = img_dict["width"]
+        if "expressions" in img_dict:
+            record["expressions"] = img_dict["expressions"]
+        image_id = record["image_id"] = img_dict["id"]
+
+        objs = []
+        for anno in anno_dict_list:
+            # Check that the image_id in this annotation is the same as
+            # the image_id we're looking at.
+            # This fails only when the data parsing logic or the annotation file is buggy.
+
+            # The original COCO valminusminival2014 & minival2014 annotation files
+            # actually contains bugs that, together with certain ways of using COCO API,
+            # can trigger this assertion.
+            assert anno["image_id"] == image_id
+
+            assert anno.get("ignore", 0) == 0, '"ignore" in COCO json file is not supported.'
+
+            obj = {key: anno[key] for key in ann_keys if key in anno}
+            if "bbox" in obj and len(obj["bbox"]) == 0:
+                raise ValueError(
+                    f"One annotation of image {image_id} contains empty 'bbox' value! "
+                    "This json does not have valid COCO format."
+                )
+
+            segm = anno.get("segmentation", None)
+            if segm:  # either list[list[float]] or dict(RLE)
+                if isinstance(segm, dict):
+                    if isinstance(segm["counts"], list):
+                        # convert to compressed RLE
+                        segm = mask_util.frPyObjects(segm, *segm["size"])
+                else:
+                    # filter out invalid polygons (< 3 points)
+                    segm = [poly for poly in segm if len(poly) % 2 == 0 and len(poly) >= 6]
+                    if len(segm) == 0:
+                        num_instances_without_valid_segmentation += 1
+                        continue  # ignore this instance
+                obj["segmentation"] = segm
+
+            keypts = anno.get("keypoints", None)
+            if keypts:  # list[int]
+                for idx, v in enumerate(keypts):
+                    if idx % 3 != 2:
+                        # COCO's segmentation coordinates are floating points in [0, H or W],
+                        # but keypoint coordinates are integers in [0, H-1 or W-1]
+                        # Therefore we assume the coordinates are "pixel indices" and
+                        # add 0.5 to convert to floating point coordinates.
+                        keypts[idx] = v + 0.5
+                obj["keypoints"] = keypts
+
+            phrase = anno.get("phrase", None)
+            if phrase:
+                obj["phrase"] = phrase
+            obj["bbox_mode"] = BoxMode.XYWH_ABS
+            if id_map:
+                annotation_category_id = obj["category_id"]
+                try:
+                    obj["category_id"] = id_map[annotation_category_id]
+                except KeyError as e:
+                    raise KeyError(
+                        f"Encountered category_id={annotation_category_id} "
+                        "but this id does not exist in 'categories' of the json file."
+                    ) from e
+            objs.append(obj)
+        record["annotations"] = objs
+        record["task"] = "grounding"
+        dataset_dicts.append(record)
+
+    if num_instances_without_valid_segmentation > 0:
+        logger.warning(
+            "Filtered out {} instances without valid segmentation. ".format(
+                num_instances_without_valid_segmentation
+            )
+            + "There might be issues in your dataset generation process.  Please "
+            "check https://detectron2.readthedocs.io/en/latest/tutorials/datasets.html carefully"
+        )
+    return dataset_dicts
+
+
+def register_refcoco(name, metadata, json_file, image_root):
+    """
+    Register a dataset in COCO's json annotation format for
+    instance detection, instance segmentation and keypoint detection.
+    (i.e., Type 1 and 2 in http://cocodataset.org/#format-data.
+    `instances*.json` and `person_keypoints*.json` in the dataset).
+
+    This is an example of how to register a new dataset.
+    You can do something similar to this function, to register new datasets.
+
+    Args:
+        name (str): the name that identifies a dataset, e.g. "coco_2014_train".
+        metadata (dict): extra metadata associated with this dataset.  You can
+            leave it as an empty dict.
+        json_file (str): path to the json instance annotation file.
+        image_root (str or path-like): directory which contains all the images.
+    """
+    assert isinstance(name, str), name
+    assert isinstance(json_file, (str, os.PathLike)), json_file
+    assert isinstance(image_root, (str, os.PathLike)), image_root
+    # 1. register a function which returns dicts
+    DatasetCatalog.register(name, lambda: load_refcoco_json(json_file, image_root, name))
+
+    # 2. Optionally, add metadata about this dataset,
+    # since they might be useful in evaluation, visualization or logging
+    MetadataCatalog.get(name).set(
+        json_file=json_file, image_root=image_root, evaluator_type="refcoco", **metadata
+    )
+
+
+# ==== Predefined splits for REFCOCO datasets ===========
+_PREDEFINED_SPLITS_REFCOCO = {
+    # refcoco
+    "refcoco-unc-train": ("coco/train2014", "SeqTR/refcoco-unc/instances_cocofied_train.json"),
+    "refcoco-unc-val": ("coco/train2014", "SeqTR/refcoco-unc/instances_cocofied_val.json"),
+    "refcoco-unc-testA": ("coco/train2014", "SeqTR/refcoco-unc/instances_cocofied_testA.json"),
+    "refcoco-unc-testB": ("coco/train2014", "SeqTR/refcoco-unc/instances_cocofied_testB.json"),
+    # refcocog
+    "refcocog-umd-train": ("coco/train2014", "SeqTR/refcocog-umd/instances_cocofied_train.json"),
+    "refcocog-umd-val": ("coco/train2014", "SeqTR/refcocog-umd/instances_cocofied_val.json"),
+    "refcocog-umd-test": ("coco/train2014", "SeqTR/refcocog-umd/instances_cocofied_test.json"),
+    "refcocog-google-val": ("coco/train2014", "SeqTR/refcocog-google/instances_cocofied_val.json"),
+    # refcoco+
+    "refcocoplus-unc-train": (
+        "coco/train2014",
+        "SeqTR/refcocoplus-unc/instances_cocofied_train.json",
+    ),
+    "refcocoplus-unc-val": ("coco/train2014", "SeqTR/refcocoplus-unc/instances_cocofied_val.json"),
+    "refcocoplus-unc-testA": (
+        "coco/train2014",
+        "SeqTR/refcocoplus-unc/instances_cocofied_testA.json",
+    ),
+    "refcocoplus-unc-testB": (
+        "coco/train2014",
+        "SeqTR/refcocoplus-unc/instances_cocofied_testB.json",
+    ),
+    # mixed
+    "refcoco-mixed": ("coco/train2014", "SeqTR/refcoco-mixed/instances_cocofied_train.json"),
+    "refcoco-mixed-filter": (
+        "coco/train2014",
+        "SeqTR/refcoco-mixed/instances_cocofied_train_filter.json",
+    ),
+    "refcoco-mixed_group-by-image": (
+        "coco/train2014",
+        "SeqTR/refcoco-mixed_group-by-image/instances_cocofied_train.json",
+    ),
+}
+
+
+def register_all_refcoco(root):
+    for key, (image_root, json_file) in _PREDEFINED_SPLITS_REFCOCO.items():
+        # Assume pre-defined datasets live in `./datasets`.
+        register_refcoco(
+            key,
+            _get_refcoco_meta(),
+            os.path.join(root, json_file) if "://" not in json_file else json_file,
+            os.path.join(root, image_root),
+        )
+
+
+if __name__.endswith(".refcoco"):
+    # Assume pre-defined datasets live in `./datasets`.
+    _root = os.getenv("DETECTRON2_DATASETS", "datasets")
+    register_all_refcoco(_root)

approach/ovod/APE/ape/data/datasets/register_bdd100k_panoseg.py

@@ -0,0 +1,277 @@
+# --------------------------------------------------------
+# X-Decoder -- Generalized Decoding for Pixel, Image, and Language
+# Copyright (c) 2022 Microsoft
+# Licensed under The MIT License [see LICENSE for details]
+# Modified by Xueyan Zou (xueyan@cs.wisc.edu)
+# --------------------------------------------------------
+# Copyright (c) Facebook, Inc. and its affiliates.
+import json
+import os
+from collections import namedtuple
+
+from detectron2.data import DatasetCatalog, MetadataCatalog
+from detectron2.utils.file_io import PathManager
+
+Label = namedtuple(
+    "Label",
+    [
+        "name",  # The identifier of this label, e.g. 'car', 'person', ... .
+        # We use them to uniquely name a class
+        "id",  # An integer ID that is associated with this label.
+        # The IDs are used to represent the label in ground truth images An ID
+        # of -1 means that this label does not have an ID and thus is ignored
+        # when creating ground truth images (e.g. license plate). Do not modify
+        # these IDs, since exactly these IDs are expected by the evaluation
+        # server.
+        "trainId",
+        # Feel free to modify these IDs as suitable for your method. Then
+        # create ground truth images with train IDs, using the tools provided
+        # in the 'preparation' folder. However, make sure to validate or submit
+        # results to our evaluation server using the regular IDs above! For
+        # trainIds, multiple labels might have the same ID. Then, these labels
+        # are mapped to the same class in the ground truth images. For the
+        # inverse mapping, we use the label that is defined first in the list
+        # below. For example, mapping all void-type classes to the same ID in
+        # training, might make sense for some approaches. Max value is 255!
+        "category",  # The name of the category that this label belongs to
+        "categoryId",
+        # The ID of this category. Used to create ground truth images
+        # on category level.
+        "hasInstances",
+        # Whether this label distinguishes between single instances or not
+        "ignoreInEval",
+        # Whether pixels having this class as ground truth label are ignored
+        # during evaluations or not
+        "color",  # The color of this label
+    ],
+)
+
+
+# Our extended list of label types. Our train id is compatible with Cityscapes
+BDD_CATEGORIES = [
+    #       name                     id    trainId   category catId
+    #       hasInstances   ignoreInEval   color
+    # Label("unlabeled", 0, 255, "void", 0, False, True, (0, 0, 0)),
+    Label("dynamic", 1, 255, "void", 0, False, True, (111, 74, 0)),
+    Label("ego vehicle", 2, 255, "void", 0, False, True, (0, 0, 0)),
+    Label("ground", 3, 255, "void", 0, False, True, (81, 0, 81)),
+    Label("static", 4, 255, "void", 0, False, True, (0, 0, 0)),
+    Label("parking", 5, 255, "flat", 1, False, True, (250, 170, 160)),
+    Label("rail track", 6, 255, "flat", 1, False, True, (230, 150, 140)),
+    Label("road", 7, 0, "flat", 1, False, False, (128, 64, 128)),
+    Label("sidewalk", 8, 1, "flat", 1, False, False, (244, 35, 232)),
+    Label("bridge", 9, 255, "construction", 2, False, True, (150, 100, 100)),
+    Label("building", 10, 2, "construction", 2, False, False, (70, 70, 70)),
+    Label("fence", 11, 4, "construction", 2, False, False, (190, 153, 153)),
+    Label("garage", 12, 255, "construction", 2, False, True, (180, 100, 180)),
+    Label("guard rail", 13, 255, "construction", 2, False, True, (180, 165, 180)),
+    Label("tunnel", 14, 255, "construction", 2, False, True, (150, 120, 90)),
+    Label("wall", 15, 3, "construction", 2, False, False, (102, 102, 156)),
+    Label("banner", 16, 255, "object", 3, False, True, (250, 170, 100)),
+    Label("billboard", 17, 255, "object", 3, False, True, (220, 220, 250)),
+    Label("lane divider", 18, 255, "object", 3, False, True, (255, 165, 0)),
+    Label("parking sign", 19, 255, "object", 3, False, False, (220, 20, 60)),
+    Label("pole", 20, 5, "object", 3, False, False, (153, 153, 153)),
+    Label("polegroup", 21, 255, "object", 3, False, True, (153, 153, 153)),
+    Label("street light", 22, 255, "object", 3, False, True, (220, 220, 100)),
+    Label("traffic cone", 23, 255, "object", 3, False, True, (255, 70, 0)),
+    Label("traffic device", 24, 255, "object", 3, False, True, (220, 220, 220)),
+    Label("traffic light", 25, 6, "object", 3, False, False, (250, 170, 30)),
+    Label("traffic sign", 26, 7, "object", 3, False, False, (220, 220, 0)),
+    Label(
+        "traffic sign frame",
+        27,
+        255,
+        "object",
+        3,
+        False,
+        True,
+        (250, 170, 250),
+    ),
+    Label("terrain", 28, 9, "nature", 4, False, False, (152, 251, 152)),
+    Label("vegetation", 29, 8, "nature", 4, False, False, (107, 142, 35)),
+    Label("sky", 30, 10, "sky", 5, False, False, (70, 130, 180)),
+    Label("person", 31, 11, "human", 6, True, False, (220, 20, 60)),
+    Label("rider", 32, 12, "human", 6, True, False, (255, 0, 0)),
+    Label("bicycle", 33, 18, "vehicle", 7, True, False, (119, 11, 32)),
+    Label("bus", 34, 15, "vehicle", 7, True, False, (0, 60, 100)),
+    Label("car", 35, 13, "vehicle", 7, True, False, (0, 0, 142)),
+    Label("caravan", 36, 255, "vehicle", 7, True, True, (0, 0, 90)),
+    Label("motorcycle", 37, 17, "vehicle", 7, True, False, (0, 0, 230)),
+    Label("trailer", 38, 255, "vehicle", 7, True, True, (0, 0, 110)),
+    Label("train", 39, 16, "vehicle", 7, True, False, (0, 80, 100)),
+    Label("truck", 40, 14, "vehicle", 7, True, False, (0, 0, 70)),
+]
+
+BDD_COLORS = [k.color for k in BDD_CATEGORIES]
+
+MetadataCatalog.get("bdd100k_pano_val").set(
+    stuff_colors=BDD_COLORS[:],
+)
+
+
+def load_bdd_panoptic_json(json_file, image_dir, gt_dir, meta):
+    """
+    Args:
+        image_dir (str): path to the raw dataset. e.g., "~/coco/train2017".
+        gt_dir (str): path to the raw annotations. e.g., "~/coco/panoptic_train2017".
+        json_file (str): path to the json file. e.g., "~/coco/annotations/panoptic_train2017.json".
+    Returns:
+        list[dict]: a list of dicts in Detectron2 standard format. (See
+        `Using Custom Datasets </tutorials/datasets.html>`_ )
+    """
+
+    def _convert_category_id(segment_info, meta):
+        if segment_info["category_id"] in meta["thing_dataset_id_to_contiguous_id"]:
+            segment_info["category_id"] = meta["thing_dataset_id_to_contiguous_id"][
+                segment_info["category_id"]
+            ]
+            segment_info["isthing"] = True
+        else:
+            segment_info["category_id"] = meta["stuff_dataset_id_to_contiguous_id"][
+                segment_info["category_id"]
+            ]
+            segment_info["isthing"] = False
+        return segment_info
+
+    with PathManager.open(json_file) as f:
+        json_info = json.load(f)
+
+    ret = []
+    for ann in json_info["annotations"]:
+        image_id = ann["image_id"]
+        # TODO: currently we assume image and label has the same filename but
+        # different extension, and images have extension ".jpg" for COCO. Need
+        # to make image extension a user-provided argument if we extend this
+        # function to support other COCO-like datasets.
+        file_name = ann["file_name"].replace("png", "jpg")
+
+        image_file = os.path.join(image_dir, file_name)
+        label_file = os.path.join(gt_dir, ann["file_name"])
+
+        segments_info = [_convert_category_id(x, meta) for x in ann["segments_info"]]
+        ret.append(
+            {
+                "file_name": image_file,
+                "image_id": image_id,
+                "pan_seg_file_name": label_file,
+                "segments_info": segments_info,
+            }
+        )
+    assert len(ret), f"No images found in {image_dir}!"
+    assert PathManager.isfile(ret[0]["file_name"]), ret[0]["file_name"]
+    assert PathManager.isfile(ret[0]["pan_seg_file_name"]), ret[0]["pan_seg_file_name"]
+    return ret
+
+
+def register_bdd_panoptic(
+    name,
+    metadata,
+    image_root,
+    panoptic_root,
+    panoptic_json,
+):
+    """
+    Register a "standard" version of ADE20k panoptic segmentation dataset named `name`.
+    The dictionaries in this registered dataset follows detectron2's standard format.
+    Hence it's called "standard".
+    Args:
+        name (str): the name that identifies a dataset,
+            e.g. "ade20k_panoptic_train"
+        metadata (dict): extra metadata associated with this dataset.
+        image_root (str): directory which contains all the images
+        panoptic_root (str): directory which contains panoptic annotation images in COCO format
+        panoptic_json (str): path to the json panoptic annotation file in COCO format
+        sem_seg_root (none): not used, to be consistent with
+            `register_coco_panoptic_separated`.
+        instances_json (str): path to the json instance annotation file
+    """
+    panoptic_name = name
+    DatasetCatalog.register(
+        panoptic_name,
+        lambda: load_bdd_panoptic_json(panoptic_json, image_root, panoptic_root, metadata),
+    )
+    MetadataCatalog.get(panoptic_name).set(
+        panoptic_root=panoptic_root,
+        image_root=image_root,
+        panoptic_json=panoptic_json,
+        evaluator_type="bdd_panoptic_pano",
+        ignore_label=0,
+        label_divisor=1000,
+        **metadata,
+    )
+
+
+_PREDEFINED_SPLITS_SCANNET_PANOPTIC = {
+    "bdd10k_40_panoptic_val": (
+        "bdd100k/images/10k/val",
+        "bdd100k/labels/pan_seg/coco_pano/val",
+        "bdd100k/labels/pan_seg/meta/coco_val.json",
+    ),
+}
+
+
+def get_metadata():
+    meta = {}
+    # The following metadata maps contiguous id from [0, #thing categories +
+    # #stuff categories) to their names and colors. We have to replica of the
+    # same name and color under "thing_*" and "stuff_*" because the current
+    # visualization function in D2 handles thing and class classes differently
+    # due to some heuristic used in Panoptic FPN. We keep the same naming to
+    # enable reusing existing visualization functions.
+    thing_classes = [k.name for k in BDD_CATEGORIES if k.hasInstances == True]
+    thing_colors = [k.color for k in BDD_CATEGORIES if k.hasInstances == True]
+    stuff_classes = [k.name for k in BDD_CATEGORIES]
+    stuff_colors = [k.color for k in BDD_CATEGORIES]
+
+    meta["thing_classes"] = thing_classes
+    meta["thing_colors"] = thing_colors
+    meta["stuff_classes"] = stuff_classes
+    meta["stuff_colors"] = stuff_colors
+
+    # Convert category id for training:
+    #   category id: like semantic segmentation, it is the class id for each
+    #   pixel. Since there are some classes not used in evaluation, the category
+    #   id is not always contiguous and thus we have two set of category ids:
+    #       - original category id: category id in the original dataset, mainly
+    #           used for evaluation.
+    #       - contiguous category id: [0, #classes), in order to train the linear
+    #           softmax classifier.
+    thing_dataset_id_to_contiguous_id = {}
+    stuff_dataset_id_to_contiguous_id = {}
+
+    for i, cat in enumerate(BDD_CATEGORIES):
+        if cat.hasInstances:
+            thing_dataset_id_to_contiguous_id[cat.id] = i
+        # else:
+        #     stuff_dataset_id_to_contiguous_id[cat["id"]] = i
+
+        # in order to use sem_seg evaluator
+        stuff_dataset_id_to_contiguous_id[cat.id] = i
+
+    meta["thing_dataset_id_to_contiguous_id"] = thing_dataset_id_to_contiguous_id
+    meta["stuff_dataset_id_to_contiguous_id"] = stuff_dataset_id_to_contiguous_id
+    return meta
+
+
+def register_all_bdd_panoptic(root):
+    metadata = get_metadata()
+    for (
+        prefix,
+        (image_root, panoptic_root, panoptic_json),
+    ) in _PREDEFINED_SPLITS_SCANNET_PANOPTIC.items():
+        # The "standard" version of COCO panoptic segmentation dataset,
+        # e.g. used by Panoptic-DeepLab
+        register_bdd_panoptic(
+            prefix,
+            metadata,
+            os.path.join(root, image_root),
+            os.path.join(root, panoptic_root),
+            os.path.join(root, panoptic_json),
+        )
+
+
+if __name__.endswith(".register_bdd100k_panoseg"):
+    # Assume pre-defined datasets live in `./datasets`.
+    _root = os.getenv("DETECTRON2_DATASETS", "datasets")
+    register_all_bdd_panoptic(_root)

approach/ovod/APE/ape/data/datasets/register_bdd100k_semseg.py

@@ -0,0 +1,98 @@
+# --------------------------------------------------------
+# X-Decoder -- Generalized Decoding for Pixel, Image, and Language
+# Copyright (c) 2022 Microsoft
+# Licensed under The MIT License [see LICENSE for details]
+# Modified by Xueyan Zou (xueyan@cs.wisc.edu)
+# --------------------------------------------------------
+# Copyright (c) Facebook, Inc. and its affiliates.
+import glob
+import os
+from typing import List, Tuple, Union
+
+import numpy as np
+
+from detectron2.data import DatasetCatalog, MetadataCatalog
+from detectron2.utils.file_io import PathManager
+
+BDD_SEM = [
+    "road",
+    "sidewalk",
+    "building",
+    "wall",
+    "fence",
+    "pole",
+    "traffic light",
+    "traffic sign",
+    "vegetation",
+    "terrain",
+    "sky",
+    "person",
+    "rider",
+    "car",
+    "truck",
+    "bus",
+    "train",
+    "motorcycle",
+    "bicycle",
+]
+
+__all__ = ["load_scannet_instances", "register_scannet_context"]
+
+
+def load_bdd_instances(
+    name: str, dirname: str, split: str, class_names: Union[List[str], Tuple[str, ...]]
+):
+    """
+    Load BDD annotations to Detectron2 format.
+
+    Args:
+        dirname: Contain "Annotations", "ImageSets", "JPEGImages"
+        split (str): one of "train", "test", "val", "trainval"
+        class_names: list or tuple of class names
+    """
+    img_folder = os.path.join(dirname, "images", "10k", split)
+    img_pths = sorted(glob.glob(os.path.join(img_folder, "*.jpg")))
+
+    sem_folder = os.path.join(dirname, "labels", "sem_seg", "masks", split)
+    sem_pths = sorted(glob.glob(os.path.join(sem_folder, "*.png")))
+
+    assert len(img_pths) == len(sem_pths)
+
+    dicts = []
+    for img_pth, sem_pth in zip(img_pths, sem_pths):
+        r = {
+            "file_name": img_pth,
+            "sem_seg_file_name": sem_pth,
+            "image_id": img_pth.split("/")[-1].split(".")[0],
+        }
+        dicts.append(r)
+    return dicts
+
+
+def register_bdd_context(name, dirname, split, class_names=BDD_SEM):
+    DatasetCatalog.register(name, lambda: load_bdd_instances(name, dirname, split, class_names))
+    MetadataCatalog.get(name).set(
+        stuff_classes=class_names,
+        dirname=dirname,
+        split=split,
+        ignore_label=[255],
+        thing_dataset_id_to_contiguous_id={},
+        class_offset=0,
+        keep_sem_bgd=False,
+    )
+
+
+def register_all_bdd_semseg(root):
+    SPLITS = [
+        ("bdd10k_val_sem_seg", "bdd100k", "val"),
+    ]
+
+    for name, dirname, split in SPLITS:
+        register_bdd_context(name, os.path.join(root, dirname), split)
+        MetadataCatalog.get(name).evaluator_type = "sem_seg"
+
+
+if __name__.endswith(".register_bdd100k_semseg"):
+    # Assume pre-defined datasets live in `./datasets`.
+    _root = os.getenv("DATASET", "datasets")
+    register_all_bdd_semseg(_root)

approach/ovod/APE/ape/data/datasets/register_pascal_context.py

@@ -0,0 +1,587 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+import os
+
+from detectron2.data import DatasetCatalog, MetadataCatalog
+from detectron2.data.datasets import load_sem_seg
+
+PASCALCONTEX59_NAMES = (
+    "aeroplane",
+    "bicycle",
+    "bird",
+    "boat",
+    "bottle",
+    "bus",
+    "car",
+    "cat",
+    "chair",
+    "cow",
+    "table",
+    "dog",
+    "horse",
+    "motorbike",
+    "person",
+    "pottedplant",
+    "sheep",
+    "sofa",
+    "train",
+    "tvmonitor",
+    "bag",
+    "bed",
+    "bench",
+    "book",
+    "building",
+    "cabinet",
+    "ceiling",
+    "cloth",
+    "computer",
+    "cup",
+    "door",
+    "fence",
+    "floor",
+    "flower",
+    "food",
+    "grass",
+    "ground",
+    "keyboard",
+    "light",
+    "mountain",
+    "mouse",
+    "curtain",
+    "platform",
+    "sign",
+    "plate",
+    "road",
+    "rock",
+    "shelves",
+    "sidewalk",
+    "sky",
+    "snow",
+    "bedclothes",
+    "track",
+    "tree",
+    "truck",
+    "wall",
+    "water",
+    "window",
+    "wood",
+)
+
+PASCALCONTEX459_NAMES = (
+    "accordion",
+    "aeroplane",
+    "air conditioner",
+    "antenna",
+    "artillery",
+    "ashtray",
+    "atrium",
+    "baby carriage",
+    "bag",
+    "ball",
+    "balloon",
+    "bamboo weaving",
+    "barrel",
+    "baseball bat",
+    "basket",
+    "basketball backboard",
+    "bathtub",
+    "bed",
+    "bedclothes",
+    "beer",
+    "bell",
+    "bench",
+    "bicycle",
+    "binoculars",
+    "bird",
+    "bird cage",
+    "bird feeder",
+    "bird nest",
+    "blackboard",
+    "board",
+    "boat",
+    "bone",
+    "book",
+    "bottle",
+    "bottle opener",
+    "bowl",
+    "box",
+    "bracelet",
+    "brick",
+    "bridge",
+    "broom",
+    "brush",
+    "bucket",
+    "building",
+    "bus",
+    "cabinet",
+    "cabinet door",
+    "cage",
+    "cake",
+    "calculator",
+    "calendar",
+    "camel",
+    "camera",
+    "camera lens",
+    "can",
+    "candle",
+    "candle holder",
+    "cap",
+    "car",
+    "card",
+    "cart",
+    "case",
+    "casette recorder",
+    "cash register",
+    "cat",
+    "cd",
+    "cd player",
+    "ceiling",
+    "cell phone",
+    "cello",
+    "chain",
+    "chair",
+    "chessboard",
+    "chicken",
+    "chopstick",
+    "clip",
+    "clippers",
+    "clock",
+    "closet",
+    "cloth",
+    "clothes tree",
+    "coffee",
+    "coffee machine",
+    "comb",
+    "computer",
+    "concrete",
+    "cone",
+    "container",
+    "control booth",
+    "controller",
+    "cooker",
+    "copying machine",
+    "coral",
+    "cork",
+    "corkscrew",
+    "counter",
+    "court",
+    "cow",
+    "crabstick",
+    "crane",
+    "crate",
+    "cross",
+    "crutch",
+    "cup",
+    "curtain",
+    "cushion",
+    "cutting board",
+    "dais",
+    "disc",
+    "disc case",
+    "dishwasher",
+    "dock",
+    "dog",
+    "dolphin",
+    "door",
+    "drainer",
+    "dray",
+    "drink dispenser",
+    "drinking machine",
+    "drop",
+    "drug",
+    "drum",
+    "drum kit",
+    "duck",
+    "dumbbell",
+    "earphone",
+    "earrings",
+    "egg",
+    "electric fan",
+    "electric iron",
+    "electric pot",
+    "electric saw",
+    "electronic keyboard",
+    "engine",
+    "envelope",
+    "equipment",
+    "escalator",
+    "exhibition booth",
+    "extinguisher",
+    "eyeglass",
+    "fan",
+    "faucet",
+    "fax machine",
+    "fence",
+    "ferris wheel",
+    "fire extinguisher",
+    "fire hydrant",
+    "fire place",
+    "fish",
+    "fish tank",
+    "fishbowl",
+    "fishing net",
+    "fishing pole",
+    "flag",
+    "flagstaff",
+    "flame",
+    "flashlight",
+    "floor",
+    "flower",
+    "fly",
+    "foam",
+    "food",
+    "footbridge",
+    "forceps",
+    "fork",
+    "forklift",
+    "fountain",
+    "fox",
+    "frame",
+    "fridge",
+    "frog",
+    "fruit",
+    "funnel",
+    "furnace",
+    "game controller",
+    "game machine",
+    "gas cylinder",
+    "gas hood",
+    "gas stove",
+    "gift box",
+    "glass",
+    "glass marble",
+    "globe",
+    "glove",
+    "goal",
+    "grandstand",
+    "grass",
+    "gravestone",
+    "ground",
+    "guardrail",
+    "guitar",
+    "gun",
+    "hammer",
+    "hand cart",
+    "handle",
+    "handrail",
+    "hanger",
+    "hard disk drive",
+    "hat",
+    "hay",
+    "headphone",
+    "heater",
+    "helicopter",
+    "helmet",
+    "holder",
+    "hook",
+    "horse",
+    "horse-drawn carriage",
+    "hot-air balloon",
+    "hydrovalve",
+    "ice",
+    "inflator pump",
+    "ipod",
+    "iron",
+    "ironing board",
+    "jar",
+    "kart",
+    "kettle",
+    "key",
+    "keyboard",
+    "kitchen range",
+    "kite",
+    "knife",
+    "knife block",
+    "ladder",
+    "ladder truck",
+    "ladle",
+    "laptop",
+    "leaves",
+    "lid",
+    "life buoy",
+    "light",
+    "light bulb",
+    "lighter",
+    "line",
+    "lion",
+    "lobster",
+    "lock",
+    "machine",
+    "mailbox",
+    "mannequin",
+    "map",
+    "mask",
+    "mat",
+    "match book",
+    "mattress",
+    "menu",
+    "metal",
+    "meter box",
+    "microphone",
+    "microwave",
+    "mirror",
+    "missile",
+    "model",
+    "money",
+    "monkey",
+    "mop",
+    "motorbike",
+    "mountain",
+    "mouse",
+    "mouse pad",
+    "musical instrument",
+    "napkin",
+    "net",
+    "newspaper",
+    "oar",
+    "ornament",
+    "outlet",
+    "oven",
+    "oxygen bottle",
+    "pack",
+    "pan",
+    "paper",
+    "paper box",
+    "paper cutter",
+    "parachute",
+    "parasol",
+    "parterre",
+    "patio",
+    "pelage",
+    "pen",
+    "pen container",
+    "pencil",
+    "person",
+    "photo",
+    "piano",
+    "picture",
+    "pig",
+    "pillar",
+    "pillow",
+    "pipe",
+    "pitcher",
+    "plant",
+    "plastic",
+    "plate",
+    "platform",
+    "player",
+    "playground",
+    "pliers",
+    "plume",
+    "poker",
+    "poker chip",
+    "pole",
+    "pool table",
+    "postcard",
+    "poster",
+    "pot",
+    "pottedplant",
+    "printer",
+    "projector",
+    "pumpkin",
+    "rabbit",
+    "racket",
+    "radiator",
+    "radio",
+    "rail",
+    "rake",
+    "ramp",
+    "range hood",
+    "receiver",
+    "recorder",
+    "recreational machines",
+    "remote control",
+    "road",
+    "robot",
+    "rock",
+    "rocket",
+    "rocking horse",
+    "rope",
+    "rug",
+    "ruler",
+    "runway",
+    "saddle",
+    "sand",
+    "saw",
+    "scale",
+    "scanner",
+    "scissors",
+    "scoop",
+    "screen",
+    "screwdriver",
+    "sculpture",
+    "scythe",
+    "sewer",
+    "sewing machine",
+    "shed",
+    "sheep",
+    "shell",
+    "shelves",
+    "shoe",
+    "shopping cart",
+    "shovel",
+    "sidecar",
+    "sidewalk",
+    "sign",
+    "signal light",
+    "sink",
+    "skateboard",
+    "ski",
+    "sky",
+    "sled",
+    "slippers",
+    "smoke",
+    "snail",
+    "snake",
+    "snow",
+    "snowmobiles",
+    "sofa",
+    "spanner",
+    "spatula",
+    "speaker",
+    "speed bump",
+    "spice container",
+    "spoon",
+    "sprayer",
+    "squirrel",
+    "stage",
+    "stair",
+    "stapler",
+    "stick",
+    "sticky note",
+    "stone",
+    "stool",
+    "stove",
+    "straw",
+    "stretcher",
+    "sun",
+    "sunglass",
+    "sunshade",
+    "surveillance camera",
+    "swan",
+    "sweeper",
+    "swim ring",
+    "swimming pool",
+    "swing",
+    "switch",
+    "table",
+    "tableware",
+    "tank",
+    "tap",
+    "tape",
+    "tarp",
+    "telephone",
+    "telephone booth",
+    "tent",
+    "tire",
+    "toaster",
+    "toilet",
+    "tong",
+    "tool",
+    "toothbrush",
+    "towel",
+    "toy",
+    "toy car",
+    "track",
+    "train",
+    "trampoline",
+    "trash bin",
+    "tray",
+    "tree",
+    "tricycle",
+    "tripod",
+    "trophy",
+    "truck",
+    "tube",
+    "turtle",
+    "tvmonitor",
+    "tweezers",
+    "typewriter",
+    "umbrella",
+    "unknown",
+    "vacuum cleaner",
+    "vending machine",
+    "video camera",
+    "video game console",
+    "video player",
+    "video tape",
+    "violin",
+    "wakeboard",
+    "wall",
+    "wallet",
+    "wardrobe",
+    "washing machine",
+    "watch",
+    "water",
+    "water dispenser",
+    "water pipe",
+    "water skate board",
+    "watermelon",
+    "whale",
+    "wharf",
+    "wheel",
+    "wheelchair",
+    "window",
+    "window blinds",
+    "wineglass",
+    "wire",
+    "wood",
+    "wool",
+)
+
+
+def _get_voc_meta(cat_list):
+    ret = {
+        "stuff_classes": cat_list,
+    }
+    return ret
+
+
+def register_pascal_context_59(root):
+    root = os.path.join(root, "VOCdevkit/VOC2010")
+    meta = _get_voc_meta(PASCALCONTEX59_NAMES)
+    for name, image_dirname, sem_seg_dirname in [
+        ("val", "JPEGImages", "annotations_detectron2/pc59_val"),
+    ]:
+        image_dir = os.path.join(root, image_dirname)
+        gt_dir = os.path.join(root, sem_seg_dirname)
+        all_name = f"pascal_context_59_sem_seg_{name}"
+        DatasetCatalog.register(
+            all_name,
+            lambda x=image_dir, y=gt_dir: load_sem_seg(y, x, gt_ext="png", image_ext="jpg"),
+        )
+        MetadataCatalog.get(all_name).set(
+            image_root=image_dir,
+            sem_seg_root=gt_dir,
+            evaluator_type="sem_seg",
+            ignore_label=255,
+            **meta,
+        )
+
+
+def register_pascal_context_459(root):
+    root = os.path.join(root, "VOCdevkit/VOC2010")
+    meta = _get_voc_meta(PASCALCONTEX459_NAMES)
+    for name, image_dirname, sem_seg_dirname in [
+        ("val", "JPEGImages", "annotations_detectron2/pc459_val"),
+    ]:
+        image_dir = os.path.join(root, image_dirname)
+        gt_dir = os.path.join(root, sem_seg_dirname)
+        all_name = f"pascal_context_459_sem_seg_{name}"
+        DatasetCatalog.register(
+            all_name,
+            lambda x=image_dir, y=gt_dir: load_sem_seg(y, x, gt_ext="tif", image_ext="jpg"),
+        )
+        MetadataCatalog.get(all_name).set(
+            image_root=image_dir,
+            sem_seg_root=gt_dir,
+            evaluator_type="sem_seg",
+            ignore_label=65535,  # NOTE: gt is saved in 16-bit TIFF images
+            **meta,
+        )
+
+
+if __name__.endswith(".register_pascal_context"):
+    # Assume pre-defined datasets live in `./datasets`.
+    _root = os.getenv("DETECTRON2_DATASETS", "datasets")
+    register_pascal_context_59(_root)
+    register_pascal_context_459(_root)

approach/ovod/APE/ape/data/datasets/register_voc_seg.py

@@ -0,0 +1,64 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+import os
+
+from detectron2.data import DatasetCatalog, MetadataCatalog
+from detectron2.data.datasets import load_sem_seg
+
+PASCALVOC20_NAMES = (
+    "aeroplane",
+    "bicycle",
+    "bird",
+    "boat",
+    "bottle",
+    "bus",
+    "car",
+    "cat",
+    "chair",
+    "cow",
+    "diningtable",
+    "dog",
+    "horse",
+    "motorbike",
+    "person",
+    "pottedplant",
+    "sheep",
+    "sofa",
+    "train",
+    "tvmonitor",
+)
+
+
+def _get_voc_meta(cat_list):
+    ret = {
+        "stuff_classes": cat_list,
+    }
+    return ret
+
+
+def register_pascalvoc(root):
+    root = os.path.join(root, "VOCdevkit/VOC2012")
+    meta = _get_voc_meta(PASCALVOC20_NAMES)
+
+    for name, image_dirname, sem_seg_dirname in [
+        ("val", "JPEGImages", "annotations_detectron2/val"),
+    ]:
+        image_dir = os.path.join(root, image_dirname)
+        gt_dir = os.path.join(root, sem_seg_dirname)
+        all_name = f"pascalvoc20_sem_seg_{name}"
+        DatasetCatalog.register(
+            all_name,
+            lambda x=image_dir, y=gt_dir: load_sem_seg(y, x, gt_ext="png", image_ext="jpg"),
+        )
+        MetadataCatalog.get(all_name).set(
+            image_root=image_dir,
+            sem_seg_root=gt_dir,
+            evaluator_type="sem_seg",
+            ignore_label=255,
+            **meta,
+        )
+
+
+if __name__.endswith(".register_voc_seg"):
+    # Assume pre-defined datasets live in `./datasets`.
+    _root = os.getenv("DETECTRON2_DATASETS", "datasets")
+    register_pascalvoc(_root)

approach/ovod/APE/ape/data/datasets/sa1b.py

@@ -0,0 +1,44 @@
+import os
+
+from detectron2.data.datasets.register_coco import register_coco_instances
+
+SA1B_CATEGORIES = [
+    {"id": 1, "name": "object"},
+]
+
+
+def _get_builtin_metadata(key):
+    id_to_name = {x["id"]: x["name"] for x in SA1B_CATEGORIES}
+    thing_dataset_id_to_contiguous_id = {i + 1: i for i in range(len(SA1B_CATEGORIES))}
+    thing_classes = [id_to_name[k] for k in sorted(id_to_name)]
+    return {
+        "thing_dataset_id_to_contiguous_id": thing_dataset_id_to_contiguous_id,
+        "thing_classes": thing_classes,
+    }
+
+
+_PREDEFINED_SPLITS_SA1B = {
+    "sa1b": ("SA-1B/images", "SA-1B/sam1b_instance.json"),
+    "sa1b_1m": ("SA-1B/images", "SA-1B/sam1b_instance_1000000.json"),
+    "sa1b_2m": ("SA-1B/images", "SA-1B/sam1b_instance_2000000.json"),
+    "sa1b_4m": ("SA-1B/images", "SA-1B/sam1b_instance_4000000.json"),
+    "sa1b_6m": ("SA-1B/images", "SA-1B/sam1b_instance_6000000.json"),
+    "sa1b_8m": ("SA-1B/images", "SA-1B/sam1b_instance_8000000.json"),
+    "sa1b_10m": ("SA-1B/images", "SA-1B/sam1b_instance_10000000.json"),
+}
+
+
+def register_all_sa1b(root):
+    for key, (image_root, json_file) in _PREDEFINED_SPLITS_SA1B.items():
+        register_coco_instances(
+            key,
+            _get_builtin_metadata(key),
+            os.path.join(root, json_file) if "://" not in json_file else json_file,
+            os.path.join(root, image_root),
+        )
+
+
+if __name__.endswith(".sa1b"):
+    # Assume pre-defined datasets live in `./datasets`.
+    _root = os.getenv("DETECTRON2_DATASETS", "datasets")
+    register_all_sa1b(_root)

approach/ovod/APE/ape/data/datasets/seginw_categories.py

@@ -0,0 +1,89 @@
+SEGINW_CATEGORIES = {
+    "seginw_Helmet-Head": ["Helmet"],
+    "seginw_Line-Contour": ["line-structure"],
+    "seginw_Elephants": ["elephant"],
+    "seginw_Hand-Metal": ["hand", "metal"],
+    "seginw_Watermelon": ["watermelon"],
+    "seginw_House-Parts": [
+        "aluminium door",
+        "aluminium window",
+        "cellar window",
+        "mint cond roof",
+        "plaster",
+        "plastic door",
+        "plastic window",
+        "plate fascade",
+        "wooden door",
+        "wooden fascade",
+        "wooden window",
+        "worn cond roof",
+    ],
+    "seginw_HouseHold-Items": ["bottle", "mouse", "perfume", "phone"],
+    "seginw_Strawberry": ["R_strawberry", "people"],
+    "seginw_WareHouse-Box": ["box", "dmg box", "label"],
+    "seginw_Fruits": ["apple", "lemon", "orange", "pear", "strawberry"],
+    "seginw_Nutterfly-Squireel": ["butterfly", "squirrel"],
+    "seginw_Hand": ["Hand-Segmentation", "hand"],
+    "seginw_Garbage": ["bin", "garbage", "pavement", "road"],
+    "seginw_Chicken": ["chicken"],
+    "seginw_Rail": ["rail"],
+    "seginw_Airplane-Parts": ["Airplane", "Body", "Cockpit", "Engine", "Wing"],
+    "seginw_Face-Mask": ["Mask"],
+    "seginw_Brain-Tumor": ["tumor"],
+    "seginw_Poles": ["poles"],
+    "seginw_Car-Parts": [
+        "back_bumper",
+        "back_door",
+        "back_glass",
+        "back_light",
+        "front_bumper",
+        "front_door",
+        "front_glass",
+        "front_light",
+        "hood",
+    ],
+    "seginw_Electric-Shaver": ["caorau"],
+    "seginw_Bottles": ["bottle", "can", "label"],
+    "seginw_Toolkits": [
+        "Allen-key",
+        "block",
+        "gasket",
+        "plier",
+        "prism",
+        "screw",
+        "screwdriver",
+        "wrench",
+    ],
+    "seginw_Trash": [
+        "Aluminium foil",
+        "Cigarette",
+        "Clear plastic bottle",
+        "Corrugated carton",
+        "Disposable plastic cup",
+        "Drink Can",
+        "Egg Carton",
+        "Foam cup",
+        "Food Can",
+        "Garbage bag",
+        "Glass bottle",
+        "Glass cup",
+        "Metal bottle cap",
+        "Other carton",
+        "Other plastic bottle",
+        "Paper cup",
+        "Plastic bag - wrapper",
+        "Plastic bottle cap",
+        "Plastic lid",
+        "Plastic straw",
+        "Pop tab",
+        "Styrofoam piece",
+    ],
+    "seginw_Salmon-Fillet": ["Salmon_fillet"],
+    "seginw_Puppies": ["puppy"],
+    "seginw_Tablets": ["tablets"],
+    "seginw_Cable": ["cable"],
+    "seginw_Fire": ["fire"],
+    "seginw_Phones": ["phone"],
+    "seginw_Cows": ["cow"],
+    "seginw_Ginger-Garlic": ["garlic", "ginger"],
+}

approach/ovod/APE/ape/data/datasets/seginw_instance.py

@@ -0,0 +1,142 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+import collections
+import json
+import os
+
+from detectron2.data import DatasetCatalog, MetadataCatalog
+from detectron2.data.datasets import load_sem_seg
+from detectron2.data.datasets.builtin_meta import COCO_CATEGORIES
+from detectron2.utils.file_io import PathManager
+
+from .seginw_categories import SEGINW_CATEGORIES
+
+_CATEGORIES = [
+    "Elephants",
+    "Hand-Metal",
+    "Watermelon",
+    "House-Parts",
+    "HouseHold-Items",
+    "Strawberry",
+    "Fruits",
+    "Nutterfly-Squireel",
+    "Hand",
+    "Garbage",
+    "Chicken",
+    "Rail",
+    "Airplane-Parts",
+    "Brain-Tumor",
+    "Poles",
+    "Electric-Shaver",
+    "Bottles",
+    "Toolkits",
+    "Trash",
+    "Salmon-Fillet",
+    "Puppies",
+    "Tablets",
+    "Phones",
+    "Cows",
+    "Ginger-Garlic",
+]
+
+_PREDEFINED_SPLITS_SEGINW = {
+    "seginw_{}_val".format(cat): (
+        "valid",
+        "seginw/{}".format(cat),  # image_root
+        "_annotations_min1cat.coco.json",  # annot_root
+    )
+    for cat in _CATEGORIES
+}
+_PREDEFINED_SPLITS_SEGINW.update(
+    {
+        "seginw_{}_train".format(cat): (
+            "train",
+            "seginw/{}".format(cat),  # image_root
+            "_annotations_min1cat.coco.json",  # annot_root
+        )
+        for cat in _CATEGORIES
+    }
+)
+
+
+def get_metadata(name):
+    # meta = {"thing_dataset_id_to_contiguous_id": {}}
+    meta = {}
+    meta["thing_classes"] = SEGINW_CATEGORIES[name.replace("_train", "").replace("_val", "")]
+    meta["thing_dataset_id_to_contiguous_id"] = {i: i for i in range(len(meta["thing_classes"]))}
+    return meta
+
+
+def load_seginw_json(name, image_root, annot_json, metadata):
+    """
+    Args:
+        image_dir (str): path to the raw dataset. e.g., "~/coco/train2017".
+        gt_dir (str): path to the raw annotations. e.g., "~/coco/panoptic_train2017".
+        json_file (str): path to the json file. e.g., "~/coco/annotations/panoptic_train2017.json".
+    Returns:
+        list[dict]: a list of dicts in Detectron2 standard format. (See
+        `Using Custom Datasets </tutorials/datasets.html>`_ )
+    """
+
+    with PathManager.open(annot_json) as f:
+        json_info = json.load(f)
+
+    # build dictionary for grounding
+    grd_dict = collections.defaultdict(list)
+    for grd_ann in json_info["annotations"]:
+        image_id = int(grd_ann["image_id"])
+        grd_dict[image_id].append(grd_ann)
+
+    ret = []
+    for image in json_info["images"]:
+        image_id = int(image["id"])
+        image_file = os.path.join(image_root, image["file_name"])
+        grounding_anno = grd_dict[image_id]
+
+        if "train" in name and len(grounding_anno) == 0:
+            continue
+
+        ret.append(
+            {
+                "file_name": image_file,
+                "image_id": image_id,
+                "inst_info": grounding_anno,
+            }
+        )
+
+    assert len(ret), f"No images found in {image_root}!"
+    assert PathManager.isfile(ret[0]["file_name"]), ret[0]["file_name"]
+    return ret
+
+
+def register_seginw(name, metadata, image_root, annot_json):
+    DatasetCatalog.register(
+        name,
+        lambda: load_seginw_json(name, image_root, annot_json, metadata),
+    )
+    MetadataCatalog.get(name).set(
+        image_root=image_root,
+        json_file=annot_json,
+        evaluator_type="seginw",
+        ignore_label=255,
+        label_divisor=1000,
+        **metadata,
+    )
+
+
+def register_all_seginw(root):
+    for (
+        prefix,
+        (split, folder_name, annot_name),
+    ) in _PREDEFINED_SPLITS_SEGINW.items():
+        register_seginw(
+            prefix,
+            get_metadata(prefix),
+            os.path.join(root, folder_name, split),
+            os.path.join(root, folder_name, split, annot_name),
+        )
+
+
+if __name__.endswith(".seginw_instance"):
+    # Assume pre-defined datasets live in `./datasets`.
+    _root = os.getenv("DATASET", "datasets")
+    register_all_seginw(_root)

approach/ovod/APE/ape/data/datasets/visualgenome.py

@@ -0,0 +1,220 @@
+import logging
+import os
+
+from .coco import custom_register_coco_instances
+from .visualgenome_categories import (
+    VISUALGENOME_150_CATEGORIES,
+    VISUALGENOME_1356_CATEGORIES,
+    VISUALGENOME_1356MINUS150_CATEGORIES,
+    VISUALGENOME_77962_CATEGORIES,
+    VISUALGENOME_77962MINUS150_CATEGORIES,
+)
+
+logger = logging.getLogger(__name__)
+
+
+def _get_builtin_metadata(dataset_name):
+    if dataset_name == "visualgenome_150_box":
+        return _get_visualgenome_metadata(VISUALGENOME_150_CATEGORIES)
+
+    if dataset_name == "visualgenome_region":
+        return _get_visualgenome_metadata([])
+
+    if dataset_name == "visualgenome_150_box_and_region":
+        return _get_visualgenome_metadata(VISUALGENOME_150_CATEGORIES)
+
+    if dataset_name == "visualgenome_77962_box_and_region":
+        return _get_visualgenome_metadata(VISUALGENOME_77962_CATEGORIES)
+
+    if dataset_name == "visualgenome_77962_box":
+        return _get_visualgenome_metadata(VISUALGENOME_77962_CATEGORIES)
+
+    if dataset_name == "visualgenome_77962minus150_box":
+        return _get_visualgenome_metadata(VISUALGENOME_77962MINUS150_CATEGORIES)
+
+    if dataset_name == "visualgenome_77962minus2319_box":
+        return _get_visualgenome_metadata(VISUALGENOME_77962MINUS150_CATEGORIES)
+
+    if dataset_name == "visualgenome_1356_box":
+        return _get_visualgenome_metadata(VISUALGENOME_1356_CATEGORIES)
+
+    if dataset_name == "visualgenome_1356minus150_box":
+        return _get_visualgenome_metadata(VISUALGENOME_1356MINUS150_CATEGORIES)
+
+    if dataset_name == "visualgenome_1356minus2319_box":
+        return _get_visualgenome_metadata(VISUALGENOME_1356MINUS150_CATEGORIES)
+
+    raise KeyError("No built-in metadata for dataset {}".format(dataset_name))
+
+
+def _get_visualgenome_metadata(categories):
+    if len(categories) == 0:
+        return {}
+    id_to_name = {x["id"]: x["name"] for x in categories}
+    thing_dataset_id_to_contiguous_id = {i + 1: i for i in range(len(categories))}
+    thing_classes = [id_to_name[k] for k in sorted(id_to_name)]
+    return {
+        "thing_dataset_id_to_contiguous_id": thing_dataset_id_to_contiguous_id,
+        "thing_classes": thing_classes,
+    }
+
+
+_PREDEFINED_SPLITS_VISUALGENOME = {}
+_PREDEFINED_SPLITS_VISUALGENOME["visualgenome_150_box"] = {
+    "visualgenome_150_box_train": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_150_box_train.json",
+    ),
+    "visualgenome_150_box_val": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_150_box_val.json",
+    ),
+}
+
+_PREDEFINED_SPLITS_VISUALGENOME["visualgenome_150_box_and_region"] = {
+    "visualgenome_150_box_and_region_train": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_150_box_and_region_train.json",
+    ),
+    "visualgenome_150_box_and_region_val": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_150_box_and_region_val.json",
+    ),
+}
+
+_PREDEFINED_SPLITS_VISUALGENOME["visualgenome_77962_box"] = {
+    "visualgenome_77962_box": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_77962_box.json",
+    ),
+    "visualgenome_77962_box_train": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_77962_box_train.json",
+    ),
+    "visualgenome_77962_box_val": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_77962_box_val.json",
+    ),
+}
+
+_PREDEFINED_SPLITS_VISUALGENOME["visualgenome_77962_box_and_region"] = {
+    "visualgenome_77962_box_and_region": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_77962_box_and_region.json",
+    ),
+    "visualgenome_77962_box_and_region_train": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_77962_box_and_region_train.json",
+    ),
+    "visualgenome_77962_box_and_region_val": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_77962_box_and_region_val.json",
+    ),
+}
+
+_PREDEFINED_SPLITS_VISUALGENOME["visualgenome_region"] = {
+    "visualgenome_region": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_region.json",
+    ),
+    "visualgenome_region_train": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_region_train.json",
+    ),
+    "visualgenome_region_val": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_region_val.json",
+    ),
+}
+
+_PREDEFINED_SPLITS_VISUALGENOME["visualgenome_77962minus150_box"] = {
+    "visualgenome_77962minus150_box": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_77962minus150_box.json",
+    ),
+    "visualgenome_77962minus150_box_train": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_77962minus150_box_train.json",
+    ),
+    "visualgenome_77962minus150_box_val": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_77962minus150_box_val.json",
+    ),
+}
+
+_PREDEFINED_SPLITS_VISUALGENOME["visualgenome_77962minus2319_box"] = {
+    "visualgenome_77962minus2319_box": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_77962minus2319_box.json",
+    ),
+    "visualgenome_77962minus2319_box_train": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_77962minus2319_box_train.json",
+    ),
+    "visualgenome_77962minus2319_box_val": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_77962minus2319_box_val.json",
+    ),
+}
+
+_PREDEFINED_SPLITS_VISUALGENOME["visualgenome_1356_box"] = {
+    "visualgenome_1356_box": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_1356_box.json",
+    ),
+    "visualgenome_1356_box_train": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_1356_box_train.json",
+    ),
+    "visualgenome_1356_box_val": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_1356_box_val.json",
+    ),
+}
+
+_PREDEFINED_SPLITS_VISUALGENOME["visualgenome_1356minus150_box"] = {
+    "visualgenome_1356minus150_box": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_1356minus150_box.json",
+    ),
+    "visualgenome_1356minus150_box_train": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_1356minus150_box_train.json",
+    ),
+    "visualgenome_1356minus150_box_val": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_1356minus150_box_val.json",
+    ),
+}
+
+_PREDEFINED_SPLITS_VISUALGENOME["visualgenome_1356minus2319_box"] = {
+    "visualgenome_1356minus2319_box": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_1356minus2319_box.json",
+    ),
+    "visualgenome_1356minus2319_box_train": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_1356minus2319_box_train.json",
+    ),
+    "visualgenome_1356minus2319_box_val": (
+        "visualgenome",
+        "visualgenome/annotations/visualgenome_1356minus2319_box_val.json",
+    ),
+}
+
+
+def register_all_visualgenome(root):
+    for dataset_name, splits_per_dataset in _PREDEFINED_SPLITS_VISUALGENOME.items():
+        for key, (image_root, json_file) in splits_per_dataset.items():
+            custom_register_coco_instances(
+                key,
+                _get_builtin_metadata(dataset_name),
+                os.path.join(root, json_file) if "://" not in json_file else json_file,
+                os.path.join(root, image_root),
+            )
+
+
+if __name__.endswith(".visualgenome"):
+    # Assume pre-defined datasets live in `./datasets`.
+    _root = os.path.expanduser(os.getenv("DETECTRON2_DATASETS", "datasets"))
+    register_all_visualgenome(_root)

approach/ovod/APE/ape/data/samplers/__init__.py

@@ -0,0 +1,5 @@
+from .distributed_sampler_multi_dataset import MultiDatasetTrainingSampler
+
+__all__ = [
+    "MultiDatasetTrainingSampler",
+]

approach/ovod/APE/ape/data/samplers/distributed_sampler_multi_dataset.py

@@ -0,0 +1,137 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+import itertools
+import logging
+import math
+from collections import defaultdict
+from typing import Optional
+
+import torch
+from torch.utils.data.sampler import Sampler
+
+from detectron2.data.samplers import RepeatFactorTrainingSampler
+from detectron2.utils import comm
+
+logger = logging.getLogger(__name__)
+
+
+class MultiDatasetTrainingSampler(Sampler):
+    def __init__(self, repeat_factors, *, shuffle=True, seed=None):
+        self._shuffle = shuffle
+        if seed is None:
+            seed = comm.shared_random_seed()
+        self._seed = int(seed)
+
+        self._rank = comm.get_rank()
+        self._world_size = comm.get_world_size()
+
+        # Split into whole number (_int_part) and fractional (_frac_part) parts.
+        self._int_part = torch.trunc(repeat_factors)
+        self._frac_part = repeat_factors - self._int_part
+
+    @staticmethod
+    def get_repeat_factors(
+        dataset_dicts, num_datasets, dataset_ratio, use_rfs, use_cas, repeat_thresh, cas_lambda
+    ):
+        sizes = [0 for _ in range(num_datasets)]
+        for d in dataset_dicts:
+            sizes[d["dataset_id"]] += 1
+
+        assert len(dataset_ratio) == len(
+            sizes
+        ), "length of dataset ratio {} should be equal to number if dataset {}".format(
+            len(dataset_ratio), len(sizes)
+        )
+        dataset_weight = [
+            torch.ones(s, dtype=torch.float32) * max(sizes) / s * r
+            for i, (r, s) in enumerate(zip(dataset_ratio, sizes))
+        ]
+
+        logger = logging.getLogger(__name__)
+        logger.info(
+            "Training sampler dataset weight: {}".format(
+                str([max(sizes) / s * r for i, (r, s) in enumerate(zip(dataset_ratio, sizes))])
+            )
+        )
+
+        st = 0
+        repeat_factors = []
+        for i, s in enumerate(sizes):
+            assert use_rfs[i] * use_cas[i] == 0
+            if use_rfs[i]:
+                repeat_factor = RepeatFactorTrainingSampler.repeat_factors_from_category_frequency(
+                    dataset_dicts[st : st + s], repeat_thresh
+                )
+            elif use_cas[i]:
+                repeat_factor = MultiDatasetTrainingSampler.get_class_balance_factor_per_dataset(
+                    dataset_dicts[st : st + s], l=cas_lambda
+                )
+                repeat_factor = repeat_factor * (s / repeat_factor.sum())
+            else:
+                repeat_factor = torch.ones(s, dtype=torch.float32)
+            logger.info(
+                "Training sampler class weight: {} {} {}".format(
+                    repeat_factor.size(), repeat_factor.max(), repeat_factor.min()
+                )
+            )
+            repeat_factors.append(repeat_factor)
+            st = st + s
+        repeat_factors = torch.cat(repeat_factors)
+        dataset_weight = torch.cat(dataset_weight)
+        repeat_factors = dataset_weight * repeat_factors
+
+        return repeat_factors
+
+    @staticmethod
+    def get_class_balance_factor_per_dataset(dataset_dicts, l=1.0):
+        rep_factors = []
+        category_freq = defaultdict(int)
+        for dataset_dict in dataset_dicts:  # For each image (without repeats)
+            cat_ids = {ann["category_id"] for ann in dataset_dict["annotations"]}
+            for cat_id in cat_ids:
+                category_freq[cat_id] += 1
+        for dataset_dict in dataset_dicts:
+            cat_ids = {ann["category_id"] for ann in dataset_dict["annotations"]}
+            rep_factor = sum([1.0 / (category_freq[cat_id] ** l) for cat_id in cat_ids])
+            rep_factors.append(rep_factor)
+
+        return torch.tensor(rep_factors, dtype=torch.float32)
+
+    def _get_epoch_indices(self, generator):
+        """
+        Create a list of dataset indices (with repeats) to use for one epoch.
+
+        Args:
+            generator (torch.Generator): pseudo random number generator used for
+                stochastic rounding.
+
+        Returns:
+            torch.Tensor: list of dataset indices to use in one epoch. Each index
+                is repeated based on its calculated repeat factor.
+        """
+        # Since repeat factors are fractional, we use stochastic rounding so
+        # that the target repeat factor is achieved in expectation over the
+        # course of training
+        rands = torch.rand(len(self._frac_part), generator=generator)
+        rep_factors = self._int_part + (rands < self._frac_part).float()
+        # Construct a list of indices in which we repeat images as specified
+        indices = []
+        for dataset_index, rep_factor in enumerate(rep_factors):
+            indices.extend([dataset_index] * int(rep_factor.item()))
+        return torch.tensor(indices, dtype=torch.int64)
+
+    def __iter__(self):
+        start = self._rank
+        yield from itertools.islice(self._infinite_indices(), start, None, self._world_size)
+
+    def _infinite_indices(self):
+        g = torch.Generator()
+        g.manual_seed(self._seed)
+        while True:
+            # Sample indices with repeats determined by stochastic rounding; each
+            # "epoch" may have a slightly different size due to the rounding.
+            indices = self._get_epoch_indices(g)
+            if self._shuffle:
+                randperm = torch.randperm(len(indices), generator=g)
+                yield from indices[randperm].tolist()
+            else:
+                yield from indices.tolist()

approach/ovod/APE/ape/data/transforms/__init__.py

@@ -0,0 +1,5 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+from .augmentation_aa import *
+from .augmentation_lsj import *
+
+__all__ = [k for k in globals().keys() if not k.startswith("_")]

approach/ovod/APE/ape/data/transforms/augmentation_aa.py

@@ -0,0 +1,39 @@
+from detectron2.data import transforms as T
+from fvcore.transforms.transform import Transform, TransformList
+
+
+class AutoAugment(T.Augmentation):
+    def __init__(self, cfg):
+        super().__init__()
+        self.resize = T.AugmentationList(
+            [
+                T.ResizeShortestEdge(
+                    [480, 512, 544, 576, 608, 640, 672, 704, 736, 768, 800],
+                    1333,
+                    sample_style="choice",
+                ),
+            ]
+        )
+        self.resize_crop_resize = T.AugmentationList(
+            [
+                T.ResizeShortestEdge([400, 500, 600], 1333, sample_style="choice"),
+                T.RandomCrop("absolute_range", (384, 600)),
+                T.ResizeShortestEdge(
+                    [480, 512, 544, 576, 608, 640, 672, 704, 736, 768, 800],
+                    1333,
+                    sample_style="choice",
+                ),
+            ]
+        )
+
+    def __call__(self, aug_input) -> Transform:
+
+        do = self._rand_range(low=0.0, high=1.0)
+        if do > 0.5:
+            return self.resize(aug_input)
+        else:
+            return self.resize_crop_resize(aug_input)
+
+    def __repr__(self):
+        msgs = [str(self.resize), str(self.resize_crop_resize)]
+        return "AutoAugment[{}]".format(", ".join(msgs))

approach/ovod/APE/ape/data/transforms/augmentation_lsj.py

@@ -0,0 +1,38 @@
+from detectron2.data import transforms as T
+from fvcore.transforms.transform import Transform, TransformList
+
+
+class LargeScaleJitter(T.Augmentation):
+    def __init__(self, cfg):
+        super().__init__()
+
+        image_size = cfg.INPUT.LSJ.IMAGE_SIZE
+        min_scale = cfg.INPUT.LSJ.MIN_SCALE
+        max_scale = cfg.INPUT.LSJ.MAX_SCALE
+        # pad_value = 128.0
+        pad_value = 1.0 * sum(cfg.MODEL.PIXEL_MEAN) / len(cfg.MODEL.PIXEL_MEAN)
+        seg_pad_value = cfg.INPUT.SEG_PAD_VALUE
+
+        self.resize_crop = T.AugmentationList(
+            [
+                T.ResizeScale(
+                    min_scale=min_scale,
+                    max_scale=max_scale,
+                    target_height=image_size,
+                    target_width=image_size,
+                ),
+                T.FixedSizeCrop(
+                    crop_size=(image_size, image_size),
+                    pad_value=pad_value,
+                    seg_pad_value=seg_pad_value,
+                ),
+            ]
+        )
+
+    def __call__(self, aug_input) -> Transform:
+
+        return self.resize_crop(aug_input)
+
+    def __repr__(self):
+        msgs = str(self.resize_crop)
+        return "LargeScaleJitter[{}]".format(msgs)

approach/ovod/APE/ape/modeling/backbone/utils_eva.py

@@ -0,0 +1,222 @@
+# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved
+import math
+import numpy as np
+from scipy import interpolate
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+__all__ = [
+    "window_partition",
+    "window_unpartition",
+    "add_decomposed_rel_pos",
+    "get_abs_pos",
+    "PatchEmbed",
+]
+
+
+def window_partition(x, window_size):
+    """
+    Partition into non-overlapping windows with padding if needed.
+    Args:
+        x (tensor): input tokens with [B, H, W, C].
+        window_size (int): window size.
+
+    Returns:
+        windows: windows after partition with [B * num_windows, window_size, window_size, C].
+        (Hp, Wp): padded height and width before partition
+    """
+    B, H, W, C = x.shape
+
+    pad_h = (window_size - H % window_size) % window_size
+    pad_w = (window_size - W % window_size) % window_size
+    if pad_h > 0 or pad_w > 0:
+        x = F.pad(x, (0, 0, 0, pad_w, 0, pad_h))
+    Hp, Wp = H + pad_h, W + pad_w
+
+    x = x.view(B, Hp // window_size, window_size, Wp // window_size, window_size, C)
+    windows = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(-1, window_size, window_size, C)
+    return windows, (Hp, Wp)
+
+
+def window_unpartition(windows, window_size, pad_hw, hw):
+    """
+    Window unpartition into original sequences and removing padding.
+    Args:
+        x (tensor): input tokens with [B * num_windows, window_size, window_size, C].
+        window_size (int): window size.
+        pad_hw (Tuple): padded height and width (Hp, Wp).
+        hw (Tuple): original height and width (H, W) before padding.
+
+    Returns:
+        x: unpartitioned sequences with [B, H, W, C].
+    """
+    Hp, Wp = pad_hw
+    H, W = hw
+    B = windows.shape[0] // (Hp * Wp // window_size // window_size)
+    x = windows.view(B, Hp // window_size, Wp // window_size, window_size, window_size, -1)
+    x = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(B, Hp, Wp, -1)
+
+    if Hp > H or Wp > W:
+        x = x[:, :H, :W, :].contiguous()
+    return x
+
+
+def get_rel_pos(q_size, k_size, rel_pos, interp_type):
+    """
+    Get relative positional embeddings according to the relative positions of
+        query and key sizes.
+    Args:
+        q_size (int): size of query q.
+        k_size (int): size of key k.
+        rel_pos (Tensor): relative position embeddings (L, C).
+
+    Returns:
+        Extracted positional embeddings according to relative positions.
+    """
+    max_rel_dist = int(2 * max(q_size, k_size) - 1)
+    # Interpolate rel pos if needed.
+    if rel_pos.shape[0] != max_rel_dist:
+        if interp_type == "vitdet":
+            # the vitdet impl: 
+            # https://github.com/facebookresearch/detectron2/blob/96c752ce821a3340e27edd51c28a00665dd32a30/detectron2/modeling/backbone/utils.py#L77.
+
+            rel_pos_resized = F.interpolate(
+                rel_pos.reshape(1, rel_pos.shape[0], -1).permute(0, 2, 1),
+                size=max_rel_dist,
+                mode="linear",
+            )
+            rel_pos_resized = rel_pos_resized.reshape(-1, max_rel_dist).permute(1, 0)
+        elif interp_type == "beit":
+            # steal from beit https://github.com/microsoft/unilm/tree/master/beit
+            # modified by Yuxin Fang
+
+            src_size = rel_pos.shape[0]
+            dst_size = max_rel_dist
+
+            q = 1.0903078
+            dis = []
+
+            cur = 1
+            for i in range(src_size // 2):
+                dis.append(cur)
+                cur += q ** (i + 1)
+
+            r_ids = [-_ for _ in reversed(dis)]
+            x = r_ids + [0] + dis
+            t = dst_size // 2.0
+            dx = np.arange(-t, t + 0.1, 1.0)
+
+            all_rel_pos_bias = []
+            for i in range(rel_pos.shape[1]):
+                # a hack from https://github.com/baaivision/EVA/issues/8,
+                # could also be used in fine-tuning but the performance haven't been tested.
+                z = rel_pos[:, i].view(src_size).cpu().float().detach().numpy()
+                f = interpolate.interp1d(x, z, kind='cubic', fill_value="extrapolate")
+                all_rel_pos_bias.append(
+                    torch.Tensor(f(dx)).contiguous().view(-1, 1).to(rel_pos.device))
+            rel_pos_resized = torch.cat(all_rel_pos_bias, dim=-1)
+        else:
+            raise NotImplementedError()
+    else:
+        rel_pos_resized = rel_pos
+
+    # Scale the coords with short length if shapes for q and k are different.
+    q_coords = torch.arange(q_size)[:, None] * max(k_size / q_size, 1.0)
+    k_coords = torch.arange(k_size)[None, :] * max(q_size / k_size, 1.0)
+    relative_coords = (q_coords - k_coords) + (k_size - 1) * max(q_size / k_size, 1.0)
+
+    return rel_pos_resized[relative_coords.long()]
+
+
+def add_decomposed_rel_pos(attn, q, rel_pos_h, rel_pos_w, q_size, k_size, interp_type):
+    """
+    Calculate decomposed Relative Positional Embeddings from :paper:`mvitv2`.
+    https://github.com/facebookresearch/mvit/blob/19786631e330df9f3622e5402b4a419a263a2c80/mvit/models/attention.py   # noqa B950
+    Args:
+        attn (Tensor): attention map.
+        q (Tensor): query q in the attention layer with shape (B, q_h * q_w, C).
+        rel_pos_h (Tensor): relative position embeddings (Lh, C) for height axis.
+        rel_pos_w (Tensor): relative position embeddings (Lw, C) for width axis.
+        q_size (Tuple): spatial sequence size of query q with (q_h, q_w).
+        k_size (Tuple): spatial sequence size of key k with (k_h, k_w).
+
+    Returns:
+        attn (Tensor): attention map with added relative positional embeddings.
+    """
+    q_h, q_w = q_size
+    k_h, k_w = k_size
+    Rh = get_rel_pos(q_h, k_h, rel_pos_h, interp_type)
+    Rw = get_rel_pos(q_w, k_w, rel_pos_w, interp_type)
+
+    B, _, dim = q.shape
+    r_q = q.reshape(B, q_h, q_w, dim)
+    rel_h = torch.einsum("bhwc,hkc->bhwk", r_q, Rh)
+    rel_w = torch.einsum("bhwc,wkc->bhwk", r_q, Rw)
+
+    attn = (
+        attn.view(B, q_h, q_w, k_h, k_w) + rel_h[:, :, :, :, None] + rel_w[:, :, :, None, :]
+    ).view(B, q_h * q_w, k_h * k_w)
+
+    return attn
+
+
+def get_abs_pos(abs_pos, has_cls_token, hw):
+    """
+    Calculate absolute positional embeddings. If needed, resize embeddings and remove cls_token
+        dimension for the original embeddings.
+    Args:
+        abs_pos (Tensor): absolute positional embeddings with (1, num_position, C).
+        has_cls_token (bool): If true, has 1 embedding in abs_pos for cls token.
+        hw (Tuple): size of input image tokens.
+
+    Returns:
+        Absolute positional embeddings after processing with shape (1, H, W, C)
+    """
+    h, w = hw
+    if has_cls_token:
+        abs_pos = abs_pos[:, 1:]
+    xy_num = abs_pos.shape[1]
+    size = int(math.sqrt(xy_num))
+    assert size * size == xy_num
+
+    if size != h or size != w:
+        new_abs_pos = F.interpolate(
+            abs_pos.reshape(1, size, size, -1).permute(0, 3, 1, 2),
+            size=(h, w),
+            mode="bicubic",
+            align_corners=False,
+        )
+
+        return new_abs_pos.permute(0, 2, 3, 1)
+    else:
+        return abs_pos.reshape(1, h, w, -1)
+
+
+class PatchEmbed(nn.Module):
+    """
+    Image to Patch Embedding.
+    """
+
+    def __init__(
+        self, kernel_size=(16, 16), stride=(16, 16), padding=(0, 0), in_chans=3, embed_dim=768
+    ):
+        """
+        Args:
+            kernel_size (Tuple): kernel size of the projection layer.
+            stride (Tuple): stride of the projection layer.
+            padding (Tuple): padding size of the projection layer.
+            in_chans (int): Number of input image channels.
+            embed_dim (int):  embed_dim (int): Patch embedding dimension.
+        """
+        super().__init__()
+
+        self.proj = nn.Conv2d(
+            in_chans, embed_dim, kernel_size=kernel_size, stride=stride, padding=padding
+        )
+
+    def forward(self, x):
+        x = self.proj(x)
+        # B C H W -> B H W C
+        x = x.permute(0, 2, 3, 1)
+        return x

approach/ovod/APE/ape/modeling/backbone/utils_eva02.py

@@ -0,0 +1,347 @@
+# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved
+import math
+import numpy as np
+from scipy import interpolate
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+__all__ = [
+    "window_partition",
+    "window_unpartition",
+    "add_decomposed_rel_pos",
+    "get_abs_pos",
+    "PatchEmbed",
+    "VisionRotaryEmbeddingFast",
+]
+
+
+def window_partition(x, window_size):
+    """
+    Partition into non-overlapping windows with padding if needed.
+    Args:
+        x (tensor): input tokens with [B, H, W, C].
+        window_size (int): window size.
+
+    Returns:
+        windows: windows after partition with [B * num_windows, window_size, window_size, C].
+        (Hp, Wp): padded height and width before partition
+    """
+    B, H, W, C = x.shape
+
+    pad_h = (window_size - H % window_size) % window_size
+    pad_w = (window_size - W % window_size) % window_size
+    if pad_h > 0 or pad_w > 0:
+        x = F.pad(x, (0, 0, 0, pad_w, 0, pad_h))
+    Hp, Wp = H + pad_h, W + pad_w
+
+    x = x.view(B, Hp // window_size, window_size, Wp // window_size, window_size, C)
+    windows = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(-1, window_size, window_size, C)
+    return windows, (Hp, Wp)
+
+
+def window_unpartition(windows, window_size, pad_hw, hw):
+    """
+    Window unpartition into original sequences and removing padding.
+    Args:
+        x (tensor): input tokens with [B * num_windows, window_size, window_size, C].
+        window_size (int): window size.
+        pad_hw (Tuple): padded height and width (Hp, Wp).
+        hw (Tuple): original height and width (H, W) before padding.
+
+    Returns:
+        x: unpartitioned sequences with [B, H, W, C].
+    """
+    Hp, Wp = pad_hw
+    H, W = hw
+    B = windows.shape[0] // (Hp * Wp // window_size // window_size)
+    x = windows.view(B, Hp // window_size, Wp // window_size, window_size, window_size, -1)
+    x = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(B, Hp, Wp, -1)
+
+    if Hp > H or Wp > W:
+        x = x[:, :H, :W, :].contiguous()
+    return x
+
+
+def get_rel_pos(q_size, k_size, rel_pos):
+    """
+    Get relative positional embeddings according to the relative positions of
+        query and key sizes.
+    Args:
+        q_size (int): size of query q.
+        k_size (int): size of key k.
+        rel_pos (Tensor): relative position embeddings (L, C).
+
+    Returns:
+        Extracted positional embeddings according to relative positions.
+    """
+    max_rel_dist = int(2 * max(q_size, k_size) - 1)
+    use_log_interpolation = True
+
+    # Interpolate rel pos if needed.
+    if rel_pos.shape[0] != max_rel_dist:
+        if not use_log_interpolation:
+            # Interpolate rel pos.
+            rel_pos_resized = F.interpolate(
+                rel_pos.reshape(1, rel_pos.shape[0], -1).permute(0, 2, 1),
+                size=max_rel_dist,
+                mode="linear",
+            )
+            rel_pos_resized = rel_pos_resized.reshape(-1, max_rel_dist).permute(1, 0)
+        else:
+            src_size = rel_pos.shape[0]
+            dst_size = max_rel_dist
+
+            # q = 1.13492
+            q = 1.0903078
+            dis = []
+
+            cur = 1
+            for i in range(src_size // 2):
+                dis.append(cur)
+                cur += q ** (i + 1)
+
+            r_ids = [-_ for _ in reversed(dis)]
+            x = r_ids + [0] + dis
+            t = dst_size // 2.0
+            dx = np.arange(-t, t + 0.1, 1.0)
+            all_rel_pos_bias = []
+            for i in range(rel_pos.shape[1]):
+                z = rel_pos[:, i].view(src_size).cpu().float().numpy()
+                f = interpolate.interp1d(x, z, kind='cubic', fill_value="extrapolate")
+                all_rel_pos_bias.append(
+                    torch.Tensor(f(dx)).contiguous().view(-1, 1).to(rel_pos.device))
+            rel_pos_resized = torch.cat(all_rel_pos_bias, dim=-1)
+    else:
+        rel_pos_resized = rel_pos
+
+    # Scale the coords with short length if shapes for q and k are different.
+    q_coords = torch.arange(q_size)[:, None] * max(k_size / q_size, 1.0)
+    k_coords = torch.arange(k_size)[None, :] * max(q_size / k_size, 1.0)
+    relative_coords = (q_coords - k_coords) + (k_size - 1) * max(q_size / k_size, 1.0)
+
+    return rel_pos_resized[relative_coords.long()]
+
+
+def add_decomposed_rel_pos(attn, q, rel_pos_h, rel_pos_w, q_size, k_size):
+    """
+    Calculate decomposed Relative Positional Embeddings from :paper:`mvitv2`.
+    https://github.com/facebookresearch/mvit/blob/19786631e330df9f3622e5402b4a419a263a2c80/mvit/models/attention.py   # noqa B950
+    Args:
+        attn (Tensor): attention map.
+        q (Tensor): query q in the attention layer with shape (B, q_h * q_w, C).
+        rel_pos_h (Tensor): relative position embeddings (Lh, C) for height axis.
+        rel_pos_w (Tensor): relative position embeddings (Lw, C) for width axis.
+        q_size (Tuple): spatial sequence size of query q with (q_h, q_w).
+        k_size (Tuple): spatial sequence size of key k with (k_h, k_w).
+
+    Returns:
+        attn (Tensor): attention map with added relative positional embeddings.
+    """
+    q_h, q_w = q_size
+    k_h, k_w = k_size
+    Rh = get_rel_pos(q_h, k_h, rel_pos_h)
+    Rw = get_rel_pos(q_w, k_w, rel_pos_w)
+
+    B, _, dim = q.shape
+    r_q = q.reshape(B, q_h, q_w, dim)
+    rel_h = torch.einsum("bhwc,hkc->bhwk", r_q, Rh)
+    rel_w = torch.einsum("bhwc,wkc->bhwk", r_q, Rw)
+
+    attn = (
+        attn.view(B, q_h, q_w, k_h, k_w) + rel_h[:, :, :, :, None] + rel_w[:, :, :, None, :]
+    ).view(B, q_h * q_w, k_h * k_w)
+
+    return attn
+
+
+def get_abs_pos(abs_pos, has_cls_token, hw):
+    """
+    Calculate absolute positional embeddings. If needed, resize embeddings and remove cls_token
+        dimension for the original embeddings.
+    Args:
+        abs_pos (Tensor): absolute positional embeddings with (1, num_position, C).
+        has_cls_token (bool): If true, has 1 embedding in abs_pos for cls token.
+        hw (Tuple): size of input image tokens.
+
+    Returns:
+        Absolute positional embeddings after processing with shape (1, H, W, C)
+    """
+    h, w = hw
+    if has_cls_token:
+        abs_pos = abs_pos[:, 1:]
+    xy_num = abs_pos.shape[1]
+    size = int(math.sqrt(xy_num))
+    assert size * size == xy_num
+
+    if size != h or size != w:
+        new_abs_pos = F.interpolate(
+            abs_pos.reshape(1, size, size, -1).permute(0, 3, 1, 2),
+            size=(h, w),
+            mode="bicubic",
+            align_corners=False,
+        )
+
+        return new_abs_pos.permute(0, 2, 3, 1)
+    else:
+        return abs_pos.reshape(1, h, w, -1)
+
+
+class PatchEmbed(nn.Module):
+    """
+    Image to Patch Embedding.
+    """
+
+    def __init__(
+        self, kernel_size=(16, 16), stride=(16, 16), padding=(0, 0), in_chans=3, embed_dim=768
+    ):
+        """
+        Args:
+            kernel_size (Tuple): kernel size of the projection layer.
+            stride (Tuple): stride of the projection layer.
+            padding (Tuple): padding size of the projection layer.
+            in_chans (int): Number of input image channels.
+            embed_dim (int):  embed_dim (int): Patch embedding dimension.
+        """
+        super().__init__()
+
+        self.proj = nn.Conv2d(
+            in_chans, embed_dim, kernel_size=kernel_size, stride=stride, padding=padding
+        )
+
+    def forward(self, x):
+        x = self.proj(x)
+        # B C H W -> B H W C
+        x = x.permute(0, 2, 3, 1)
+        return x
+    
+
+
+
+from math import pi
+
+import torch
+from torch import nn
+
+from einops import rearrange, repeat
+
+
+
+def broadcat(tensors, dim = -1):
+    num_tensors = len(tensors)
+    shape_lens = set(list(map(lambda t: len(t.shape), tensors)))
+    assert len(shape_lens) == 1, 'tensors must all have the same number of dimensions'
+    shape_len = list(shape_lens)[0]
+    dim = (dim + shape_len) if dim < 0 else dim
+    dims = list(zip(*map(lambda t: list(t.shape), tensors)))
+    expandable_dims = [(i, val) for i, val in enumerate(dims) if i != dim]
+    assert all([*map(lambda t: len(set(t[1])) <= 2, expandable_dims)]), 'invalid dimensions for broadcastable concatentation'
+    max_dims = list(map(lambda t: (t[0], max(t[1])), expandable_dims))
+    expanded_dims = list(map(lambda t: (t[0], (t[1],) * num_tensors), max_dims))
+    expanded_dims.insert(dim, (dim, dims[dim]))
+    expandable_shapes = list(zip(*map(lambda t: t[1], expanded_dims)))
+    tensors = list(map(lambda t: t[0].expand(*t[1]), zip(tensors, expandable_shapes)))
+    return torch.cat(tensors, dim = dim)
+
+
+
+def rotate_half(x):
+    x = rearrange(x, '... (d r) -> ... d r', r = 2)
+    x1, x2 = x.unbind(dim = -1)
+    x = torch.stack((-x2, x1), dim = -1)
+    return rearrange(x, '... d r -> ... (d r)')
+
+
+
+class VisionRotaryEmbedding(nn.Module):
+    def __init__(
+        self,
+        dim,
+        pt_seq_len,
+        ft_seq_len=None,
+        custom_freqs = None,
+        freqs_for = 'lang',
+        theta = 10000,
+        max_freq = 10,
+        num_freqs = 1,
+    ):
+        super().__init__()
+        if custom_freqs:
+            freqs = custom_freqs
+        elif freqs_for == 'lang':
+            freqs = 1. / (theta ** (torch.arange(0, dim, 2)[:(dim // 2)].float() / dim))
+        elif freqs_for == 'pixel':
+            freqs = torch.linspace(1., max_freq / 2, dim // 2) * pi
+        elif freqs_for == 'constant':
+            freqs = torch.ones(num_freqs).float()
+        else:
+            raise ValueError(f'unknown modality {freqs_for}')
+
+        if ft_seq_len is None: ft_seq_len = pt_seq_len
+        t = torch.arange(ft_seq_len) / ft_seq_len * pt_seq_len
+
+        freqs_h = torch.einsum('..., f -> ... f', t, freqs)
+        freqs_h = repeat(freqs_h, '... n -> ... (n r)', r = 2)
+
+        freqs_w = torch.einsum('..., f -> ... f', t, freqs)
+        freqs_w = repeat(freqs_w, '... n -> ... (n r)', r = 2)
+
+        freqs = broadcat((freqs_h[:, None, :], freqs_w[None, :, :]), dim = -1)
+
+        self.register_buffer("freqs_cos", freqs.cos())
+        self.register_buffer("freqs_sin", freqs.sin())
+
+        print('======== shape of rope freq', self.freqs_cos.shape, '========')
+
+    def forward(self, t, start_index = 0):
+        rot_dim = self.freqs_cos.shape[-1]
+        end_index = start_index + rot_dim
+        assert rot_dim <= t.shape[-1], f'feature dimension {t.shape[-1]} is not of sufficient size to rotate in all the positions {rot_dim}'
+        t_left, t, t_right = t[..., :start_index], t[..., start_index:end_index], t[..., end_index:]
+        t = (t * self.freqs_cos) + (rotate_half(t) * self.freqs_sin)
+        return torch.cat((t_left, t, t_right), dim = -1)
+
+
+
+
+class VisionRotaryEmbeddingFast(nn.Module):
+    def __init__(
+        self,
+        dim,
+        pt_seq_len=16,
+        ft_seq_len=None,
+        custom_freqs = None,
+        freqs_for = 'lang',
+        theta = 10000,
+        max_freq = 10,
+        num_freqs = 1,
+    ):
+        super().__init__()
+        if custom_freqs:
+            freqs = custom_freqs
+        elif freqs_for == 'lang':
+            freqs = 1. / (theta ** (torch.arange(0, dim, 2)[:(dim // 2)].float() / dim))
+        elif freqs_for == 'pixel':
+            freqs = torch.linspace(1., max_freq / 2, dim // 2) * pi
+        elif freqs_for == 'constant':
+            freqs = torch.ones(num_freqs).float()
+        else:
+            raise ValueError(f'unknown modality {freqs_for}')
+
+        if ft_seq_len is None: ft_seq_len = pt_seq_len
+        t = torch.arange(ft_seq_len) / ft_seq_len * pt_seq_len
+
+        freqs = torch.einsum('..., f -> ... f', t, freqs)
+        freqs = repeat(freqs, '... n -> ... (n r)', r = 2)
+        freqs = broadcat((freqs[:, None, :], freqs[None, :, :]), dim = -1)
+
+        freqs_cos = freqs.cos().view(-1, freqs.shape[-1])
+        freqs_sin = freqs.sin().view(-1, freqs.shape[-1])
+
+        self.register_buffer("freqs_cos", freqs_cos)
+        self.register_buffer("freqs_sin", freqs_sin)
+
+        print('======== shape of rope freq', self.freqs_cos.shape, '========')
+
+    def forward(self, t): return  t * self.freqs_cos + rotate_half(t) * self.freqs_sin
+

approach/ovod/APE/ape/modeling/backbone/vit.py

@@ -0,0 +1,30 @@
+import logging
+
+logger = logging.getLogger(__name__)
+
+
+__all__ = ["get_vit_lr_decay_rate"]
+
+def get_vit_lr_decay_rate(name, lr_decay_rate=1.0, num_layers=12):
+    """
+    Calculate lr decay rate for different ViT blocks.
+    Args:
+        name (string): parameter name.
+        lr_decay_rate (float): base lr decay rate.
+        num_layers (int): number of ViT blocks.
+
+    Returns:
+        lr decay rate for the given parameter.
+    """
+    if name.startswith("model_vision."):
+        name = name[len("model_vision."):]
+
+    layer_id = num_layers + 1
+    if name.startswith("backbone"):
+        if ".pos_embed" in name or ".patch_embed" in name:
+            layer_id = 0
+        elif ".blocks." in name and ".residual." not in name:
+            layer_id = int(name[name.find(".blocks.") :].split(".")[2]) + 1
+
+    logger.info("get_vit_lr_decay_rate: name={} num_layers={} layer_id={} lr_decay_rate={}".format(name, num_layers, layer_id, lr_decay_rate ** (num_layers + 1 - layer_id)))
+    return lr_decay_rate ** (num_layers + 1 - layer_id)

approach/ovod/APE/ape/modeling/backbone/vit_eva.py

@@ -0,0 +1,644 @@
+import logging
+import math
+import fvcore.nn.weight_init as weight_init
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from torch import Tensor, Size
+from typing import Union, List
+from torch.nn.parameter import Parameter
+import numbers
+
+from detectron2.layers import CNNBlockBase, Conv2d, get_norm
+from detectron2.modeling.backbone.fpn import _assert_strides_are_log2_contiguous
+
+from fairscale.nn.checkpoint import checkpoint_wrapper
+from timm.models.layers import DropPath, Mlp, trunc_normal_
+
+from detectron2.modeling.backbone import Backbone
+from .utils_eva import (
+    PatchEmbed,
+    add_decomposed_rel_pos,
+    get_abs_pos,
+    window_partition,
+    window_unpartition,
+)
+
+logger = logging.getLogger(__name__)
+
+
+__all__ = ["ViT", "SimpleFeaturePyramid", "get_vit_lr_decay_rate"]
+
+
+_shape_t = Union[int, List[int], Size]
+
+
+# steal from beit https://github.com/microsoft/unilm/tree/master/beit
+class LayerNormWithForceFP32(nn.Module):
+    __constants__ = ['normalized_shape', 'eps', 'elementwise_affine']
+    normalized_shape: _shape_t
+    eps: float
+    elementwise_affine: bool
+
+    def __init__(self, normalized_shape: _shape_t, eps: float = 1e-5, elementwise_affine: bool = True) -> None:
+        super(LayerNormWithForceFP32, self).__init__()
+        if isinstance(normalized_shape, numbers.Integral):
+            normalized_shape = (normalized_shape,)
+        self.normalized_shape = tuple(normalized_shape)
+        self.eps = eps
+        self.elementwise_affine = elementwise_affine
+        if self.elementwise_affine:
+            self.weight = Parameter(torch.Tensor(*normalized_shape))
+            self.bias = Parameter(torch.Tensor(*normalized_shape))
+        else:
+            self.register_parameter('weight', None)
+            self.register_parameter('bias', None)
+        self.reset_parameters()
+
+    def reset_parameters(self) -> None:
+        if self.elementwise_affine:
+            nn.init.ones_(self.weight)
+            nn.init.zeros_(self.bias)
+
+    def forward(self, input: Tensor) -> Tensor:
+        return F.layer_norm(
+            input.float(), self.normalized_shape, self.weight.float(), self.bias.float(), self.eps).type_as(input)
+
+    def extra_repr(self) -> Tensor:
+        return '{normalized_shape}, eps={eps}, ' \
+               'elementwise_affine={elementwise_affine}'.format(**self.__dict__)
+
+
+class Attention(nn.Module):
+    """Multi-head Attention block with relative position embeddings."""
+
+    def __init__(
+        self,
+        dim,
+        num_heads=8,
+        qkv_bias=True,
+        beit_like_qkv_bias=False,
+        use_rel_pos=False,
+        rel_pos_zero_init=True,
+        input_size=None,
+        interp_type="vitdet",
+    ):
+        """
+        Args:
+            dim (int): Number of input channels.
+            num_heads (int): Number of attention heads.
+            qkv_bias (bool:  If True, add a learnable bias to query, key, value.
+            rel_pos (bool): If True, add relative positional embeddings to the attention map.
+            rel_pos_zero_init (bool): If True, zero initialize relative positional parameters.
+            input_size (int or None): Input resolution for calculating the relative positional
+                parameter size.
+        """
+        super().__init__()
+        self.num_heads = num_heads
+        head_dim = dim // num_heads
+        self.scale = head_dim**-0.5
+
+        self.beit_like_qkv_bias = beit_like_qkv_bias
+        if beit_like_qkv_bias:
+            self.q_bias = nn.Parameter(torch.zeros(dim))
+            self.v_bias = nn.Parameter(torch.zeros(dim))
+
+        self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)
+        self.proj = nn.Linear(dim, dim)
+
+        self.use_rel_pos = use_rel_pos
+        self.interp_type = interp_type
+        if self.use_rel_pos:
+            # initialize relative positional embeddings
+            self.rel_pos_h = nn.Parameter(torch.zeros(2 * input_size[0] - 1, head_dim))
+            self.rel_pos_w = nn.Parameter(torch.zeros(2 * input_size[1] - 1, head_dim))
+
+            if not rel_pos_zero_init:
+                trunc_normal_(self.rel_pos_h, std=0.02)
+                trunc_normal_(self.rel_pos_w, std=0.02)
+        self.qk_float = False
+
+    def forward(self, x):
+        B, H, W, _ = x.shape
+        # qkv with shape (3, B, nHead, H * W, C)
+        if self.beit_like_qkv_bias:
+            qkv_bias = torch.cat((self.q_bias, torch.zeros_like(self.v_bias, requires_grad=False), self.v_bias))
+            qkv = torch.nn.functional.linear(input=x, weight=self.qkv.weight, bias=qkv_bias)
+            qkv = qkv.reshape(B, H * W, 3, self.num_heads, -1).permute(2, 0, 3, 1, 4)
+        else:
+            qkv = self.qkv(x).reshape(B, H * W, 3, self.num_heads, -1).permute(2, 0, 3, 1, 4)
+        # q, k, v with shape (B * nHead, H * W, C)
+        q, k, v = qkv.reshape(3, B * self.num_heads, H * W, -1).unbind(0)
+
+        if self.qk_float:
+            attn = (q.float() * self.scale) @ k.float().transpose(-2, -1)
+            if self.use_rel_pos:
+                attn = add_decomposed_rel_pos(attn, q, self.rel_pos_h, self.rel_pos_w, (H, W), (H, W), self.interp_type)
+            attn = attn.softmax(dim=-1).type_as(x)
+        else:
+            attn = (q * self.scale) @ k.transpose(-2, -1)
+            if self.use_rel_pos:
+                attn = add_decomposed_rel_pos(attn, q, self.rel_pos_h, self.rel_pos_w, (H, W), (H, W), self.interp_type)
+            attn = attn.softmax(dim=-1)
+        x = (attn @ v).view(B, self.num_heads, H, W, -1).permute(0, 2, 3, 1, 4).reshape(B, H, W, -1)
+        x = self.proj(x)
+
+        return x
+
+
+class ResBottleneckBlock(CNNBlockBase):
+    """
+    The standard bottleneck residual block without the last activation layer.
+    It contains 3 conv layers with kernels 1x1, 3x3, 1x1.
+    """
+
+    def __init__(
+        self,
+        in_channels,
+        out_channels,
+        bottleneck_channels,
+        norm="LN",
+        act_layer=nn.GELU,
+    ):
+        """
+        Args:
+            in_channels (int): Number of input channels.
+            out_channels (int): Number of output channels.
+            bottleneck_channels (int): number of output channels for the 3x3
+                "bottleneck" conv layers.
+            norm (str or callable): normalization for all conv layers.
+                See :func:`layers.get_norm` for supported format.
+            act_layer (callable): activation for all conv layers.
+        """
+        super().__init__(in_channels, out_channels, 1)
+
+        self.conv1 = Conv2d(in_channels, bottleneck_channels, 1, bias=False)
+        self.norm1 = get_norm(norm, bottleneck_channels)
+        self.act1 = act_layer()
+
+        self.conv2 = Conv2d(
+            bottleneck_channels,
+            bottleneck_channels,
+            3,
+            padding=1,
+            bias=False,
+        )
+        self.norm2 = get_norm(norm, bottleneck_channels)
+        self.act2 = act_layer()
+
+        self.conv3 = Conv2d(bottleneck_channels, out_channels, 1, bias=False)
+        self.norm3 = get_norm(norm, out_channels)
+
+        for layer in [self.conv1, self.conv2, self.conv3]:
+            weight_init.c2_msra_fill(layer)
+        for layer in [self.norm1, self.norm2]:
+            layer.weight.data.fill_(1.0)
+            layer.bias.data.zero_()
+        # zero init last norm layer.
+        self.norm3.weight.data.zero_()
+        self.norm3.bias.data.zero_()
+
+    def forward(self, x):
+        out = x
+        for layer in self.children():
+            out = layer(out)
+
+        out = x + out
+        return out
+
+
+class Block(nn.Module):
+    """Transformer blocks with support of window attention and residual propagation blocks"""
+
+    def __init__(
+        self,
+        dim,
+        num_heads,
+        mlp_ratio=4.0,
+        qkv_bias=True,
+        drop_path=0.0,
+        norm_layer=LayerNormWithForceFP32,
+        act_layer=nn.GELU,
+        use_rel_pos=False,
+        rel_pos_zero_init=True,
+        window_size=0,
+        use_residual_block=False,
+        input_size=None,
+        beit_like_qkv_bias=False,
+        beit_like_gamma=False,
+        interp_type="vitdet",
+    ):
+        """
+        Args:
+            dim (int): Number of input channels.
+            num_heads (int): Number of attention heads in each ViT block.
+            mlp_ratio (float): Ratio of mlp hidden dim to embedding dim.
+            qkv_bias (bool): If True, add a learnable bias to query, key, value.
+            drop_path (float): Stochastic depth rate.
+            norm_layer (nn.Module): Normalization layer.
+            act_layer (nn.Module): Activation layer.
+            use_rel_pos (bool): If True, add relative positional embeddings to the attention map.
+            rel_pos_zero_init (bool): If True, zero initialize relative positional parameters.
+            window_size (int): Window size for window attention blocks. If it equals 0, then not
+                use window attention.
+            use_residual_block (bool): If True, use a residual block after the MLP block.
+            input_size (int or None): Input resolution for calculating the relative positional
+                parameter size.
+            beit_like_qkv_bias (bool)
+            beit_like_gamma (bool)
+        """
+        super().__init__()
+        self.norm1 = norm_layer(dim)
+        self.attn = Attention(
+            dim,
+            num_heads=num_heads,
+            qkv_bias=qkv_bias,
+            use_rel_pos=use_rel_pos,
+            rel_pos_zero_init=rel_pos_zero_init,
+            input_size=input_size if window_size == 0 else (window_size, window_size),
+            beit_like_qkv_bias=beit_like_qkv_bias,
+            interp_type=interp_type,
+        )
+
+        self.drop_path = DropPath(drop_path) if drop_path > 0.0 else nn.Identity()
+        self.norm2 = norm_layer(dim)
+        self.mlp = Mlp(in_features=dim, hidden_features=int(dim * mlp_ratio), act_layer=act_layer)
+
+        self.window_size = window_size
+
+        self.use_residual_block = use_residual_block
+        if use_residual_block:
+            # Use a residual block with bottleneck channel as dim // 2
+            self.residual = ResBottleneckBlock(
+                in_channels=dim,
+                out_channels=dim,
+                bottleneck_channels=dim // 2,
+                norm="LN",
+                act_layer=act_layer,
+            )
+
+        self.beit_like_gamma = beit_like_gamma
+        if beit_like_gamma:
+            self.gamma_1 = nn.Parameter(torch.ones((dim)), requires_grad=True)
+            self.gamma_2 = nn.Parameter(torch.ones((dim)), requires_grad=True)
+
+    def forward(self, x):
+        shortcut = x
+        x = self.norm1(x)
+        # Window partition
+        if self.window_size > 0:
+            H, W = x.shape[1], x.shape[2]
+            x, pad_hw = window_partition(x, self.window_size)
+
+        x = self.attn(x)
+        # Reverse window partition
+        if self.window_size > 0:
+            x = window_unpartition(x, self.window_size, pad_hw, (H, W))
+
+        if self.beit_like_gamma:
+            x = shortcut + self.drop_path(self.gamma_1 * x)
+            x = x + self.drop_path(self.gamma_2 * self.mlp(self.norm2(x)))
+        else:
+            x = shortcut + self.drop_path(x)
+            x = x + self.drop_path(self.mlp(self.norm2(x)))
+
+        if self.use_residual_block:
+            x = self.residual(x.permute(0, 3, 1, 2)).permute(0, 2, 3, 1)
+
+        return x
+
+
+class ViT(Backbone):
+    """
+    This module implements Vision Transformer (ViT) backbone in :paper:`vitdet`.
+    "Exploring Plain Vision Transformer Backbones for Object Detection",
+    https://arxiv.org/abs/2203.16527
+    """
+
+    def __init__(
+        self,
+        img_size=1024,
+        patch_size=16,
+        in_chans=3,
+        embed_dim=768,
+        depth=12,
+        num_heads=12,
+        mlp_ratio=4.0,
+        qkv_bias=True,
+        drop_path_rate=0.0,
+        norm_layer=LayerNormWithForceFP32,
+        act_layer=nn.GELU,
+        use_abs_pos=True,
+        use_rel_pos=False,
+        rel_pos_zero_init=True,
+        window_size=0,
+        window_block_indexes=(),
+        residual_block_indexes=(),
+        use_act_checkpoint=False,
+        pretrain_img_size=224,
+        pretrain_use_cls_token=True,
+        out_feature="last_feat",
+        beit_like_qkv_bias=True,
+        beit_like_gamma=False,
+        freeze_patch_embed=False,
+        interp_type="vitdet", 
+        frozen_stages=-1,
+    ):
+        """
+        Args:
+            img_size (int): Input image size.
+            patch_size (int): Patch size.
+            in_chans (int): Number of input image channels.
+            embed_dim (int): Patch embedding dimension.
+            depth (int): Depth of ViT.
+            num_heads (int): Number of attention heads in each ViT block.
+            mlp_ratio (float): Ratio of mlp hidden dim to embedding dim.
+            qkv_bias (bool): If True, add a learnable bias to query, key, value.
+            drop_path_rate (float): Stochastic depth rate.
+            norm_layer (nn.Module): Normalization layer.
+            act_layer (nn.Module): Activation layer.
+            use_abs_pos (bool): If True, use absolute positional embeddings.
+            use_rel_pos (bool): If True, add relative positional embeddings to the attention map.
+            rel_pos_zero_init (bool): If True, zero initialize relative positional parameters.
+            window_size (int): Window size for window attention blocks.
+            window_block_indexes (list): Indexes for blocks using window attention.
+            residual_block_indexes (list): Indexes for blocks using conv propagation.
+            use_act_checkpoint (bool): If True, use activation checkpointing.
+            pretrain_img_size (int): input image size for pretraining models.
+            pretrain_use_cls_token (bool): If True, pretrainig models use class token.
+            out_feature (str): name of the feature from the last block.
+            beit_like_qkv_bias (bool): beit_like_model that has gamma_1 and gamma_2 in blocks and qkv_bias=False
+            beit_like_gamma (bool)
+            freeze_patch_embed (bool)
+            interp_type: "vitdet" for training / fine-ting, "beit" for eval (slightly improvement at a higher res)
+        """
+        super().__init__()
+        self.pretrain_use_cls_token = pretrain_use_cls_token
+
+        self.patch_embed = PatchEmbed(
+            kernel_size=(patch_size, patch_size),
+            stride=(patch_size, patch_size),
+            in_chans=in_chans,
+            embed_dim=embed_dim,
+        )
+
+        if use_abs_pos:
+            # Initialize absolute positional embedding with pretrain image size.
+            num_patches = (pretrain_img_size // patch_size) * (pretrain_img_size // patch_size)
+            num_positions = (num_patches + 1) if pretrain_use_cls_token else num_patches
+            self.pos_embed = nn.Parameter(torch.zeros(1, num_positions, embed_dim))
+        else:
+            self.pos_embed = None
+
+        # stochastic depth decay rule
+        dpr = [x.item() for x in torch.linspace(0, drop_path_rate, depth)]
+
+        self.blocks = nn.ModuleList()
+        if beit_like_qkv_bias:
+            qkv_bias = False
+        for i in range(depth):
+            block = Block(
+                dim=embed_dim,
+                num_heads=num_heads,
+                mlp_ratio=mlp_ratio,
+                qkv_bias=qkv_bias,
+                drop_path=dpr[i],
+                norm_layer=norm_layer,
+                act_layer=act_layer,
+                use_rel_pos=use_rel_pos,
+                rel_pos_zero_init=rel_pos_zero_init,
+                window_size=window_size if i in window_block_indexes else 0,
+                use_residual_block=i in residual_block_indexes,
+                input_size=(img_size // patch_size, img_size // patch_size),
+                beit_like_qkv_bias=beit_like_qkv_bias,
+                beit_like_gamma=beit_like_gamma,
+                interp_type=interp_type,
+            )
+            if use_act_checkpoint and i > frozen_stages - 1:
+                block = checkpoint_wrapper(block)
+            self.blocks.append(block)
+
+        self._out_feature_channels = {out_feature: embed_dim}
+        self._out_feature_strides = {out_feature: patch_size}
+        self._out_features = [out_feature]
+
+        if self.pos_embed is not None:
+            nn.init.trunc_normal_(self.pos_embed, std=0.02)
+
+        self.freeze_patch_embed = freeze_patch_embed
+        self.apply(self._init_weights)
+
+        self.frozen_stages = frozen_stages
+        self._freeze_stages()
+
+    def _freeze_stages(self):
+        if self.frozen_stages >= 0:
+            self.patch_embed.eval()
+            for param in self.patch_embed.parameters():
+                param.requires_grad = False
+
+        if self.frozen_stages >= 1 and self.pos_embed is not None:
+            self.pos_embed.requires_grad = False
+
+        if self.frozen_stages >= 2:
+            for i in range(0, self.frozen_stages - 1):
+                m = self.blocks[i]
+                m.eval()
+                for name, param in m.named_parameters():
+                    vit_lr_decay_rate = get_vit_lr_decay_rate(f"backbone.net.blocks.{i}.{name}", lr_decay_rate=0.9, num_layers=len(self.blocks))
+                    logger.info(f"freeze blocks.{i}.{name} {param.size()} {vit_lr_decay_rate}")
+                    param.requires_grad = False
+
+    def _init_weights(self, m):
+        if isinstance(m, nn.Linear):
+            nn.init.trunc_normal_(m.weight, std=0.02)
+            if isinstance(m, nn.Linear) and m.bias is not None:
+                nn.init.constant_(m.bias, 0)
+        elif isinstance(m, LayerNormWithForceFP32):
+            nn.init.constant_(m.bias, 0)
+            nn.init.constant_(m.weight, 1.0)
+
+        if self.freeze_patch_embed:
+            for n, p in self.patch_embed.named_parameters():
+                p.requires_grad = False
+
+    def forward(self, x):
+        x = self.patch_embed(x)
+        if self.pos_embed is not None:
+            x = x + get_abs_pos(
+                self.pos_embed, self.pretrain_use_cls_token, (x.shape[1], x.shape[2])
+            )
+
+        for blk in self.blocks:
+            x = blk(x)
+
+        outputs = {self._out_features[0]: x.permute(0, 3, 1, 2)}
+        return outputs
+
+
+class SimpleFeaturePyramid(Backbone):
+    """
+    This module implements SimpleFeaturePyramid in :paper:`vitdet`.
+    It creates pyramid features built on top of the input feature map.
+    """
+
+    def __init__(
+        self,
+        net,
+        in_feature,
+        out_channels,
+        scale_factors,
+        top_block=None,
+        norm="LN",
+        square_pad=0,
+    ):
+        """
+        Args:
+            net (Backbone): module representing the subnetwork backbone.
+                Must be a subclass of :class:`Backbone`.
+            in_feature (str): names of the input feature maps coming
+                from the net.
+            out_channels (int): number of channels in the output feature maps.
+            scale_factors (list[float]): list of scaling factors to upsample or downsample
+                the input features for creating pyramid features.
+            top_block (nn.Module or None): if provided, an extra operation will
+                be performed on the output of the last (smallest resolution)
+                pyramid output, and the result will extend the result list. The top_block
+                further downsamples the feature map. It must have an attribute
+                "num_levels", meaning the number of extra pyramid levels added by
+                this block, and "in_feature", which is a string representing
+                its input feature (e.g., p5).
+            norm (str): the normalization to use.
+            square_pad (int): If > 0, require input images to be padded to specific square size.
+        """
+        super(SimpleFeaturePyramid, self).__init__()
+        assert isinstance(net, Backbone)
+
+        self.scale_factors = scale_factors
+
+        input_shapes = net.output_shape()
+        strides = [int(input_shapes[in_feature].stride / scale) for scale in scale_factors]
+        _assert_strides_are_log2_contiguous(strides)
+
+        dim = input_shapes[in_feature].channels
+        self.stages = []
+        use_bias = norm == ""
+        for idx, scale in enumerate(scale_factors):
+            out_dim = dim
+            if scale == 4.0:
+                layers = [
+                    nn.ConvTranspose2d(dim, dim // 2, kernel_size=2, stride=2),
+                    get_norm(norm, dim // 2),
+                    nn.GELU(),
+                    nn.ConvTranspose2d(dim // 2, dim // 4, kernel_size=2, stride=2),
+                ]
+                out_dim = dim // 4
+            elif scale == 2.0:
+                layers = [nn.ConvTranspose2d(dim, dim // 2, kernel_size=2, stride=2)]
+                out_dim = dim // 2
+            elif scale == 1.0:
+                layers = []
+            elif scale == 0.5:
+                layers = [nn.MaxPool2d(kernel_size=2, stride=2)]
+            else:
+                raise NotImplementedError(f"scale_factor={scale} is not supported yet.")
+
+            layers.extend(
+                [
+                    Conv2d(
+                        out_dim,
+                        out_channels,
+                        kernel_size=1,
+                        bias=use_bias,
+                        norm=get_norm(norm, out_channels),
+                    ),
+                    Conv2d(
+                        out_channels,
+                        out_channels,
+                        kernel_size=3,
+                        padding=1,
+                        bias=use_bias,
+                        norm=get_norm(norm, out_channels),
+                    ),
+                ]
+            )
+            layers = nn.Sequential(*layers)
+
+            stage = int(math.log2(strides[idx]))
+            self.add_module(f"simfp_{stage}", layers)
+            self.stages.append(layers)
+
+        self.net = net
+        self.in_feature = in_feature
+        self.top_block = top_block
+        # Return feature names are "p<stage>", like ["p2", "p3", ..., "p6"]
+        self._out_feature_strides = {"p{}".format(int(math.log2(s))): s for s in strides}
+        # top block output feature maps.
+        if self.top_block is not None:
+            for s in range(stage, stage + self.top_block.num_levels):
+                self._out_feature_strides["p{}".format(s + 1)] = 2 ** (s + 1)
+
+        self._out_features = list(self._out_feature_strides.keys())
+        self._out_feature_channels = {k: out_channels for k in self._out_features}
+        self._size_divisibility = strides[-1]
+        self._square_pad = square_pad
+
+    @property
+    def padding_constraints(self):
+        return {
+            "size_divisiblity": self._size_divisibility,
+            "square_size": self._square_pad,
+        }
+
+    def forward(self, x):
+        """
+        Args:
+            x: Tensor of shape (N,C,H,W). H, W must be a multiple of ``self.size_divisibility``.
+
+        Returns:
+            dict[str->Tensor]:
+                mapping from feature map name to pyramid feature map tensor
+                in high to low resolution order. Returned feature names follow the FPN
+                convention: "p<stage>", where stage has stride = 2 ** stage e.g.,
+                ["p2", "p3", ..., "p6"].
+        """
+        bottom_up_features = self.net(x)
+        features = bottom_up_features[self.in_feature]
+        results = []
+
+        for stage in self.stages:
+            results.append(stage(features))
+
+        if self.top_block is not None:
+            if self.top_block.in_feature in bottom_up_features:
+                top_block_in_feature = bottom_up_features[self.top_block.in_feature]
+            else:
+                top_block_in_feature = results[self._out_features.index(self.top_block.in_feature)]
+            results.extend(self.top_block(top_block_in_feature))
+        assert len(self._out_features) == len(results)
+        return {f: res for f, res in zip(self._out_features, results)}
+
+
+def get_vit_lr_decay_rate(name, lr_decay_rate=1.0, num_layers=12):
+    """
+    Calculate lr decay rate for different ViT blocks.
+    Args:
+        name (string): parameter name.
+        lr_decay_rate (float): base lr decay rate.
+        num_layers (int): number of ViT blocks.
+
+    Returns:
+        lr decay rate for the given parameter.
+    """
+    if name.startswith("model_vision."):
+        name = name[len("model_vision."):]
+
+    layer_id = num_layers + 1
+    if name.startswith("backbone"):
+        if ".pos_embed" in name or ".patch_embed" in name:
+            layer_id = 0
+        elif ".blocks." in name and ".residual." not in name:
+            layer_id = int(name[name.find(".blocks.") :].split(".")[2]) + 1
+
+    logger.info("get_vit_lr_decay_rate: name={} num_layers={} layer_id={} lr_decay_rate={}".format(name, num_layers, layer_id, lr_decay_rate ** (num_layers + 1 - layer_id)))
+    return lr_decay_rate ** (num_layers + 1 - layer_id)

approach/ovod/APE/ape/modeling/backbone/vit_eva02.py

@@ -0,0 +1,625 @@
+import logging
+import math
+from functools import partial
+
+import fvcore.nn.weight_init as weight_init
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+from detectron2.layers import CNNBlockBase, Conv2d, get_norm
+from detectron2.modeling.backbone.fpn import _assert_strides_are_log2_contiguous
+
+from detectron2.modeling.backbone import Backbone
+from .utils_eva02 import (
+    PatchEmbed,
+    add_decomposed_rel_pos,
+    get_abs_pos,
+    window_partition,
+    window_unpartition,
+    VisionRotaryEmbeddingFast,
+)
+
+try:
+    import xformers.ops as xops
+except:
+    pass
+
+try:
+    from apex.normalization import FusedLayerNorm
+except:
+    pass
+
+
+logger = logging.getLogger(__name__)
+
+
+
+__all__ = ["ViT", "SimpleFeaturePyramid", "get_vit_lr_decay_rate"]
+
+
+
+class SwiGLU(nn.Module):
+    def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.SiLU, drop=0., 
+                norm_layer=nn.LayerNorm, subln=False
+            ):
+        super().__init__()
+        out_features = out_features or in_features
+        hidden_features = hidden_features or in_features
+
+        self.w1 = nn.Linear(in_features, hidden_features)
+        self.w2 = nn.Linear(in_features, hidden_features)
+
+        self.act = act_layer()
+        self.ffn_ln = norm_layer(hidden_features) if subln else nn.Identity()
+        self.w3 = nn.Linear(hidden_features, out_features)
+        
+        self.drop = nn.Dropout(drop)
+
+    def forward(self, x):
+        x1 = self.w1(x)
+        x2 = self.w2(x)
+        hidden = self.act(x1) * x2
+        x = self.ffn_ln(hidden)
+        x = self.w3(x)
+        x = self.drop(x)
+        return x
+    
+
+class Attention(nn.Module):
+    def __init__(
+            self, 
+            dim, 
+            num_heads=8, 
+            qkv_bias=True, 
+            qk_scale=None, 
+            attn_head_dim=None, 
+            rope=None,
+            xattn=True,
+        ):
+        super().__init__()
+        self.num_heads = num_heads
+        head_dim = dim // num_heads
+        if attn_head_dim is not None:
+            head_dim = attn_head_dim
+        all_head_dim = head_dim * self.num_heads
+        self.scale = qk_scale or head_dim ** -0.5
+
+        self.q_proj = nn.Linear(dim, all_head_dim, bias=False)
+        self.k_proj = nn.Linear(dim, all_head_dim, bias=False)
+        self.v_proj = nn.Linear(dim, all_head_dim, bias=False)
+
+        if qkv_bias:
+            self.q_bias = nn.Parameter(torch.zeros(all_head_dim))
+            self.v_bias = nn.Parameter(torch.zeros(all_head_dim))
+        else:
+            self.q_bias = None
+            self.v_bias = None
+
+        self.rope = rope
+        self.xattn = xattn
+        self.proj = nn.Linear(all_head_dim, dim)
+
+    def forward(self, x):
+        B, H, W, C = x.shape
+        x = x.view(B, -1, C)
+        N = H * W
+
+        q = F.linear(input=x, weight=self.q_proj.weight, bias=self.q_bias)
+        k = F.linear(input=x, weight=self.k_proj.weight, bias=None)
+        v = F.linear(input=x, weight=self.v_proj.weight, bias=self.v_bias)
+
+        q = q.reshape(B, N, self.num_heads, -1).permute(0, 2, 1, 3)     # B, num_heads, N, C
+        k = k.reshape(B, N, self.num_heads, -1).permute(0, 2, 1, 3)  
+        v = v.reshape(B, N, self.num_heads, -1).permute(0, 2, 1, 3) 
+
+        ## rope
+        q = self.rope(q).type_as(v)
+        k = self.rope(k).type_as(v)
+
+        if self.xattn and not (torch.jit.is_scripting() or torch.jit.is_tracing()):
+            q = q.permute(0, 2, 1, 3)   # B, num_heads, N, C -> B, N, num_heads, C
+            k = k.permute(0, 2, 1, 3)
+            v = v.permute(0, 2, 1, 3)
+            
+            x = xops.memory_efficient_attention(q, k, v)
+            x = x.reshape(B, N, -1)
+        else:
+            q = q * self.scale
+            attn = (q @ k.transpose(-2, -1))
+            attn = attn.softmax(dim=-1).type_as(x)
+            x = (attn @ v).transpose(1, 2).reshape(B, N, -1)
+
+        x = self.proj(x)
+        x = x.view(B, H, W, C)
+
+        return x
+
+
+class ResBottleneckBlock(CNNBlockBase):
+    """
+    The standard bottleneck residual block without the last activation layer.
+    It contains 3 conv layers with kernels 1x1, 3x3, 1x1.
+    """
+
+    def __init__(
+        self,
+        in_channels,
+        out_channels,
+        bottleneck_channels,
+        norm="LN",
+        act_layer=nn.GELU,
+    ):
+        """
+        Args:
+            in_channels (int): Number of input channels.
+            out_channels (int): Number of output channels.
+            bottleneck_channels (int): number of output channels for the 3x3
+                "bottleneck" conv layers.
+            norm (str or callable): normalization for all conv layers.
+                See :func:`layers.get_norm` for supported format.
+            act_layer (callable): activation for all conv layers.
+        """
+        super().__init__(in_channels, out_channels, 1)
+
+        self.conv1 = Conv2d(in_channels, bottleneck_channels, 1, bias=False)
+        self.norm1 = get_norm(norm, bottleneck_channels)
+        self.act1 = act_layer()
+
+        self.conv2 = Conv2d(
+            bottleneck_channels,
+            bottleneck_channels,
+            3,
+            padding=1,
+            bias=False,
+        )
+        self.norm2 = get_norm(norm, bottleneck_channels)
+        self.act2 = act_layer()
+
+        self.conv3 = Conv2d(bottleneck_channels, out_channels, 1, bias=False)
+        self.norm3 = get_norm(norm, out_channels)
+
+        for layer in [self.conv1, self.conv2, self.conv3]:
+            weight_init.c2_msra_fill(layer)
+        for layer in [self.norm1, self.norm2]:
+            layer.weight.data.fill_(1.0)
+            layer.bias.data.zero_()
+        # zero init last norm layer.
+        self.norm3.weight.data.zero_()
+        self.norm3.bias.data.zero_()
+
+    def forward(self, x):
+        out = x
+        for layer in self.children():
+            out = layer(out)
+
+        out = x + out
+        return out
+
+
+class Block(nn.Module):
+    """Transformer blocks with support of window attention and residual propagation blocks"""
+
+    def __init__(
+        self,
+        dim,
+        num_heads,
+        mlp_ratio=4*2/3,
+        qkv_bias=True,
+        drop_path=0.0,
+        norm_layer=partial(nn.LayerNorm, eps=1e-6), 
+        window_size=0,
+        use_residual_block=False,
+        rope=None,
+        xattn=True,
+    ):
+        """
+        Args:
+            dim (int): Number of input channels.
+            num_heads (int): Number of attention heads in each ViT block.
+            mlp_ratio (float): Ratio of mlp hidden dim to embedding dim.
+            qkv_bias (bool): If True, add a learnable bias to query, key, value.
+            drop_path (float): Stochastic depth rate.
+            norm_layer (nn.Module): Normalization layer.
+            act_layer (nn.Module): Activation layer.
+            use_rel_pos (bool): If True, add relative positional embeddings to the attention map.
+            rel_pos_zero_init (bool): If True, zero initialize relative positional parameters.
+            window_size (int): Window size for window attention blocks. If it equals 0, then not
+                use window attention.
+            use_residual_block (bool): If True, use a residual block after the MLP block.
+            input_size (int or None): Input resolution for calculating the relative positional
+                parameter size.
+        """
+        super().__init__()
+        self.norm1 = norm_layer(dim)
+        self.attn = Attention(
+            dim,
+            num_heads=num_heads,
+            qkv_bias=qkv_bias,
+            rope=rope,
+            xattn=xattn,
+        )
+
+        from timm.models.layers import DropPath
+
+        self.drop_path = DropPath(drop_path) if drop_path > 0.0 else nn.Identity()
+        self.norm2 = norm_layer(dim)
+        self.mlp = SwiGLU(
+                in_features=dim, 
+                hidden_features=int(dim * mlp_ratio), 
+                subln=True,
+                norm_layer=norm_layer,
+            )
+
+        self.window_size = window_size
+
+        self.use_residual_block = use_residual_block
+        if use_residual_block:
+            # Use a residual block with bottleneck channel as dim // 2
+            self.residual = ResBottleneckBlock(
+                in_channels=dim,
+                out_channels=dim,
+                bottleneck_channels=dim // 2,
+                norm="LN",
+            )
+
+    def forward(self, x):
+        shortcut = x
+        x = self.norm1(x)
+
+        # Window partition
+        if self.window_size > 0:
+            H, W = x.shape[1], x.shape[2]
+            x, pad_hw = window_partition(x, self.window_size)
+
+        x = self.attn(x)
+
+        # Reverse window partition
+        if self.window_size > 0:
+            x = window_unpartition(x, self.window_size, pad_hw, (H, W))
+
+        x = shortcut + self.drop_path(x)
+        x = x + self.drop_path(self.mlp(self.norm2(x)))
+
+        if self.use_residual_block:
+            x = self.residual(x.permute(0, 3, 1, 2)).permute(0, 2, 3, 1)
+
+        return x
+
+
+class ViT(Backbone):
+    """
+    This module implements Vision Transformer (ViT) backbone in :paper:`vitdet`.
+    "Exploring Plain Vision Transformer Backbones for Object Detection",
+    https://arxiv.org/abs/2203.16527
+    """
+
+    def __init__(
+        self,
+        img_size=1024,
+        patch_size=16,
+        in_chans=3,
+        embed_dim=768,
+        depth=12,
+        num_heads=12,
+        mlp_ratio=4*2/3,
+        qkv_bias=True,
+        drop_path_rate=0.0,
+        norm_layer=partial(nn.LayerNorm, eps=1e-6),
+        act_layer=nn.GELU,
+        use_abs_pos=True,
+        use_rel_pos=False,
+        rope=True,
+        pt_hw_seq_len=16,
+        intp_freq=True,
+        window_size=0,
+        window_block_indexes=(),
+        residual_block_indexes=(),
+        use_act_checkpoint=False,
+        pretrain_img_size=224,
+        pretrain_use_cls_token=True,
+        out_feature="last_feat",
+        xattn=True,
+        frozen_stages=-1,
+    ):
+        """
+        Args:
+            img_size (int): Input image size.
+            patch_size (int): Patch size.
+            in_chans (int): Number of input image channels.
+            embed_dim (int): Patch embedding dimension.
+            depth (int): Depth of ViT.
+            num_heads (int): Number of attention heads in each ViT block.
+            mlp_ratio (float): Ratio of mlp hidden dim to embedding dim.
+            qkv_bias (bool): If True, add a learnable bias to query, key, value.
+            drop_path_rate (float): Stochastic depth rate.
+            norm_layer (nn.Module): Normalization layer.
+            act_layer (nn.Module): Activation layer.
+            use_abs_pos (bool): If True, use absolute positional embeddings.
+            use_rel_pos (bool): If True, add relative positional embeddings to the attention map.
+            rel_pos_zero_init (bool): If True, zero initialize relative positional parameters.
+            window_size (int): Window size for window attention blocks.
+            window_block_indexes (list): Indexes for blocks using window attention.
+            residual_block_indexes (list): Indexes for blocks using conv propagation.
+            use_act_checkpoint (bool): If True, use activation checkpointing.
+            pretrain_img_size (int): input image size for pretraining models.
+            pretrain_use_cls_token (bool): If True, pretrainig models use class token.
+            out_feature (str): name of the feature from the last block.
+        """
+        super().__init__()
+        self.pretrain_use_cls_token = pretrain_use_cls_token
+
+        self.patch_embed = PatchEmbed(
+            kernel_size=(patch_size, patch_size),
+            stride=(patch_size, patch_size),
+            in_chans=in_chans,
+            embed_dim=embed_dim,
+        )
+
+        if use_abs_pos:
+            # Initialize absolute positional embedding with pretrain image size.
+            num_patches = (pretrain_img_size // patch_size) * (pretrain_img_size // patch_size)
+            num_positions = (num_patches + 1) if pretrain_use_cls_token else num_patches
+            self.pos_embed = nn.Parameter(torch.zeros(1, num_positions, embed_dim))
+        else:
+            self.pos_embed = None
+
+
+        half_head_dim = embed_dim // num_heads // 2
+        hw_seq_len = img_size // patch_size
+
+        self.rope_win = VisionRotaryEmbeddingFast(
+            dim=half_head_dim,
+            pt_seq_len=pt_hw_seq_len,
+            ft_seq_len=window_size if intp_freq else None,
+        )
+        self.rope_glb = VisionRotaryEmbeddingFast(
+            dim=half_head_dim,
+            pt_seq_len=pt_hw_seq_len,
+            ft_seq_len=hw_seq_len if intp_freq else None,
+        )
+
+        # stochastic depth decay rule
+        dpr = [x.item() for x in torch.linspace(0, drop_path_rate, depth)]
+
+        self.blocks = nn.ModuleList()
+        for i in range(depth):
+            block = Block(
+                dim=embed_dim,
+                num_heads=num_heads,
+                mlp_ratio=mlp_ratio,
+                qkv_bias=qkv_bias,
+                drop_path=dpr[i],
+                norm_layer=norm_layer,
+                window_size=window_size if i in window_block_indexes else 0,
+                use_residual_block=i in residual_block_indexes,
+                rope=self.rope_win if i in window_block_indexes else self.rope_glb,
+                xattn=xattn
+            )
+            if use_act_checkpoint and i > frozen_stages - 1:
+                # TODO: use torch.utils.checkpoint
+                from fairscale.nn.checkpoint import checkpoint_wrapper
+
+                block = checkpoint_wrapper(block)
+            self.blocks.append(block)
+
+        self._out_feature_channels = {out_feature: embed_dim}
+        self._out_feature_strides = {out_feature: patch_size}
+        self._out_features = [out_feature]
+
+        if self.pos_embed is not None:
+            nn.init.trunc_normal_(self.pos_embed, std=0.02)
+
+        self.apply(self._init_weights)
+
+        self.frozen_stages = frozen_stages
+        self._freeze_stages()
+
+    def _freeze_stages(self):
+        if self.frozen_stages >= 0:
+            self.patch_embed.eval()
+            for param in self.patch_embed.parameters():
+                param.requires_grad = False
+
+        if self.frozen_stages >= 1 and self.pos_embed is not None:
+            self.pos_embed.requires_grad = False
+
+        if self.frozen_stages >= 2:
+            for i in range(0, self.frozen_stages - 1):
+                m = self.blocks[i]
+                m.eval()
+                for name, param in m.named_parameters():
+                    vit_lr_decay_rate = get_vit_lr_decay_rate(f"backbone.net.blocks.{i}.{name}", lr_decay_rate=0.9, num_layers=len(self.blocks))
+                    logger.info(f"freeze blocks.{i}.{name} {param.size()} {vit_lr_decay_rate}")
+                    param.requires_grad = False
+
+
+    def _init_weights(self, m):
+        if isinstance(m, nn.Linear):
+            nn.init.trunc_normal_(m.weight, std=0.02)
+            if isinstance(m, nn.Linear) and m.bias is not None:
+                nn.init.constant_(m.bias, 0)
+        elif isinstance(m, nn.LayerNorm):
+            nn.init.constant_(m.bias, 0)
+            nn.init.constant_(m.weight, 1.0)
+
+    def forward(self, x):
+        x = self.patch_embed(x)
+        if self.pos_embed is not None:
+            x = x + get_abs_pos(
+                self.pos_embed, self.pretrain_use_cls_token, (x.shape[1], x.shape[2])
+            )
+
+        for blk in self.blocks:
+            x = blk(x)
+
+        outputs = {self._out_features[0]: x.permute(0, 3, 1, 2)}
+        return outputs
+
+
+class SimpleFeaturePyramid(Backbone):
+    """
+    This module implements SimpleFeaturePyramid in :paper:`vitdet`.
+    It creates pyramid features built on top of the input feature map.
+    """
+
+    def __init__(
+        self,
+        net,
+        in_feature,
+        out_channels,
+        scale_factors,
+        top_block=None,
+        norm="LN",
+        square_pad=0,
+    ):
+        """
+        Args:
+            net (Backbone): module representing the subnetwork backbone.
+                Must be a subclass of :class:`Backbone`.
+            in_feature (str): names of the input feature maps coming
+                from the net.
+            out_channels (int): number of channels in the output feature maps.
+            scale_factors (list[float]): list of scaling factors to upsample or downsample
+                the input features for creating pyramid features.
+            top_block (nn.Module or None): if provided, an extra operation will
+                be performed on the output of the last (smallest resolution)
+                pyramid output, and the result will extend the result list. The top_block
+                further downsamples the feature map. It must have an attribute
+                "num_levels", meaning the number of extra pyramid levels added by
+                this block, and "in_feature", which is a string representing
+                its input feature (e.g., p5).
+            norm (str): the normalization to use.
+            square_pad (int): If > 0, require input images to be padded to specific square size.
+        """
+        super(SimpleFeaturePyramid, self).__init__()
+        assert isinstance(net, Backbone)
+
+        self.scale_factors = scale_factors
+
+        input_shapes = net.output_shape()
+        strides = [int(input_shapes[in_feature].stride / scale) for scale in scale_factors]
+        _assert_strides_are_log2_contiguous(strides)
+
+        dim = input_shapes[in_feature].channels
+        self.stages = []
+        use_bias = norm == ""
+        for idx, scale in enumerate(scale_factors):
+            out_dim = dim
+            if scale == 4.0:
+                layers = [
+                    nn.ConvTranspose2d(dim, dim // 2, kernel_size=2, stride=2),
+                    get_norm(norm, dim // 2),
+                    nn.GELU(),
+                    nn.ConvTranspose2d(dim // 2, dim // 4, kernel_size=2, stride=2),
+                ]
+                out_dim = dim // 4
+            elif scale == 2.0:
+                layers = [nn.ConvTranspose2d(dim, dim // 2, kernel_size=2, stride=2)]
+                out_dim = dim // 2
+            elif scale == 1.0:
+                layers = []
+            elif scale == 0.5:
+                layers = [nn.MaxPool2d(kernel_size=2, stride=2)]
+            else:
+                raise NotImplementedError(f"scale_factor={scale} is not supported yet.")
+
+            layers.extend(
+                [
+                    Conv2d(
+                        out_dim,
+                        out_channels,
+                        kernel_size=1,
+                        bias=use_bias,
+                        norm=get_norm(norm, out_channels),
+                    ),
+                    Conv2d(
+                        out_channels,
+                        out_channels,
+                        kernel_size=3,
+                        padding=1,
+                        bias=use_bias,
+                        norm=get_norm(norm, out_channels),
+                    ),
+                ]
+            )
+            layers = nn.Sequential(*layers)
+
+            stage = int(math.log2(strides[idx]))
+            self.add_module(f"simfp_{stage}", layers)
+            self.stages.append(layers)
+
+        self.net = net
+        self.in_feature = in_feature
+        self.top_block = top_block
+        # Return feature names are "p<stage>", like ["p2", "p3", ..., "p6"]
+        self._out_feature_strides = {"p{}".format(int(math.log2(s))): s for s in strides}
+        # top block output feature maps.
+        if self.top_block is not None:
+            for s in range(stage, stage + self.top_block.num_levels):
+                self._out_feature_strides["p{}".format(s + 1)] = 2 ** (s + 1)
+
+        self._out_features = list(self._out_feature_strides.keys())
+        self._out_feature_channels = {k: out_channels for k in self._out_features}
+        self._size_divisibility = strides[-1]
+        self._square_pad = square_pad
+
+    @property
+    def padding_constraints(self):
+        return {
+            "size_divisiblity": self._size_divisibility,
+            "square_size": self._square_pad,
+        }
+
+    def forward(self, x):
+        """
+        Args:
+            x: Tensor of shape (N,C,H,W). H, W must be a multiple of ``self.size_divisibility``.
+
+        Returns:
+            dict[str->Tensor]:
+                mapping from feature map name to pyramid feature map tensor
+                in high to low resolution order. Returned feature names follow the FPN
+                convention: "p<stage>", where stage has stride = 2 ** stage e.g.,
+                ["p2", "p3", ..., "p6"].
+        """
+        bottom_up_features = self.net(x)
+        features = bottom_up_features[self.in_feature]
+        results = []
+
+        for stage in self.stages:
+            results.append(stage(features))
+
+        if self.top_block is not None:
+            if self.top_block.in_feature in bottom_up_features:
+                top_block_in_feature = bottom_up_features[self.top_block.in_feature]
+            else:
+                top_block_in_feature = results[self._out_features.index(self.top_block.in_feature)]
+            results.extend(self.top_block(top_block_in_feature))
+        assert len(self._out_features) == len(results)
+        return {f: res for f, res in zip(self._out_features, results)}
+
+
+def get_vit_lr_decay_rate(name, lr_decay_rate=1.0, num_layers=12):
+    """
+    Calculate lr decay rate for different ViT blocks.
+    Args:
+        name (string): parameter name.
+        lr_decay_rate (float): base lr decay rate.
+        num_layers (int): number of ViT blocks.
+
+    Returns:
+        lr decay rate for the given parameter.
+    """
+    if name.startswith("model_vision."):
+        name = name[len("model_vision."):]
+
+    layer_id = num_layers + 1
+    if name.startswith("backbone"):
+        if ".pos_embed" in name or ".patch_embed" in name:
+            layer_id = 0
+        elif ".blocks." in name and ".residual." not in name:
+            layer_id = int(name[name.find(".blocks.") :].split(".")[2]) + 1
+
+    logger.info("get_vit_lr_decay_rate: name={} num_layers={} layer_id={} lr_decay_rate={}".format(name, num_layers, layer_id, lr_decay_rate ** (num_layers + 1 - layer_id)))
+    return lr_decay_rate ** (num_layers + 1 - layer_id)

approach/ovod/APE/ape/modeling/backbone/vit_eva_clip.py

@@ -0,0 +1,931 @@
+import logging
+import math
+from functools import partial
+
+import fvcore.nn.weight_init as weight_init
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+from detectron2.layers import CNNBlockBase, Conv2d, get_norm
+from detectron2.modeling.backbone.fpn import _assert_strides_are_log2_contiguous
+
+from detectron2.modeling.backbone import Backbone
+from .utils_eva02 import (
+    PatchEmbed,
+    add_decomposed_rel_pos,
+    get_abs_pos,
+    window_partition,
+    window_unpartition,
+    VisionRotaryEmbeddingFast,
+)
+
+try:
+    import xformers.ops as xops
+except:
+    pass
+
+try:
+    from apex.normalization import FusedLayerNorm
+except:
+    pass
+
+
+logger = logging.getLogger(__name__)
+
+
+
+__all__ = ["ViT", "SimpleFeaturePyramid", "get_vit_lr_decay_rate"]
+
+
+class DropPath(nn.Module):
+    """Drop paths (Stochastic Depth) per sample  (when applied in main path of residual blocks)."""
+
+    def __init__(self, drop_prob=None):
+        super(DropPath, self).__init__()
+        self.drop_prob = drop_prob
+
+    def forward(self, x):
+        return drop_path(x, self.drop_prob, self.training)
+
+    def extra_repr(self) -> str:
+        return "p={}".format(self.drop_prob)
+
+
+class Mlp(nn.Module):
+    def __init__(
+        self,
+        in_features,
+        hidden_features=None,
+        out_features=None,
+        act_layer=nn.GELU,
+        norm_layer=nn.LayerNorm,
+        drop=0.0,
+        subln=False,
+    ):
+        super().__init__()
+        out_features = out_features or in_features
+        hidden_features = hidden_features or in_features
+        self.fc1 = nn.Linear(in_features, hidden_features)
+        self.act = act_layer()
+
+        self.ffn_ln = norm_layer(hidden_features) if subln else nn.Identity()
+
+        self.fc2 = nn.Linear(hidden_features, out_features)
+        self.drop = nn.Dropout(drop)
+
+    def forward(self, x):
+        x = self.fc1(x)
+        x = self.act(x)
+        # x = self.drop(x)
+        # commit this for the orignal BERT implement
+        x = self.ffn_ln(x)
+
+        x = self.fc2(x)
+        x = self.drop(x)
+        return x
+
+
+class SwiGLU(nn.Module):
+    def __init__(
+        self,
+        in_features,
+        hidden_features=None,
+        out_features=None,
+        act_layer=nn.SiLU,
+        drop=0.0,
+        norm_layer=nn.LayerNorm,
+        subln=False,
+    ):
+        super().__init__()
+        out_features = out_features or in_features
+        hidden_features = hidden_features or in_features
+
+        self.w1 = nn.Linear(in_features, hidden_features)
+        self.w2 = nn.Linear(in_features, hidden_features)
+
+        self.act = act_layer()
+        self.ffn_ln = norm_layer(hidden_features) if subln else nn.Identity()
+        self.w3 = nn.Linear(hidden_features, out_features)
+
+        self.drop = nn.Dropout(drop)
+
+    def forward(self, x):
+        x1 = self.w1(x)
+        x2 = self.w2(x)
+        hidden = self.act(x1) * x2
+        x = self.ffn_ln(hidden)
+        x = self.w3(x)
+        x = self.drop(x)
+        return x
+
+
+class Attention(nn.Module):
+    def __init__(
+        self,
+        dim,
+        num_heads=8,
+        qkv_bias=False,
+        qk_scale=None,
+        attn_drop=0.0,
+        proj_drop=0.0,
+        window_size=0,
+        attn_head_dim=None,
+        xattn=False,
+        rope=None,
+        subln=False,
+        norm_layer=nn.LayerNorm,
+    ):
+        super().__init__()
+        self.num_heads = num_heads
+        head_dim = dim // num_heads
+        if attn_head_dim is not None:
+            head_dim = attn_head_dim
+        all_head_dim = head_dim * self.num_heads
+        self.scale = qk_scale or head_dim**-0.5
+
+        self.subln = subln
+        if self.subln:
+            self.q_proj = nn.Linear(dim, all_head_dim, bias=False)
+            self.k_proj = nn.Linear(dim, all_head_dim, bias=False)
+            self.v_proj = nn.Linear(dim, all_head_dim, bias=False)
+        else:
+            self.qkv = nn.Linear(dim, all_head_dim * 3, bias=False)
+
+        if qkv_bias:
+            self.q_bias = nn.Parameter(torch.zeros(all_head_dim))
+            self.v_bias = nn.Parameter(torch.zeros(all_head_dim))
+        else:
+            self.q_bias = None
+            self.v_bias = None
+
+        if window_size and False:
+            self.window_size = window_size
+            self.num_relative_distance = (2 * window_size[0] - 1) * (2 * window_size[1] - 1) + 3
+            self.relative_position_bias_table = nn.Parameter(
+                torch.zeros(self.num_relative_distance, num_heads)
+            )  # 2*Wh-1 * 2*Ww-1, nH
+            # cls to token & token 2 cls & cls to cls
+
+            # get pair-wise relative position index for each token inside the window
+            coords_h = torch.arange(window_size[0])
+            coords_w = torch.arange(window_size[1])
+            coords = torch.stack(torch.meshgrid([coords_h, coords_w]))  # 2, Wh, Ww
+            coords_flatten = torch.flatten(coords, 1)  # 2, Wh*Ww
+            relative_coords = (
+                coords_flatten[:, :, None] - coords_flatten[:, None, :]
+            )  # 2, Wh*Ww, Wh*Ww
+            relative_coords = relative_coords.permute(1, 2, 0).contiguous()  # Wh*Ww, Wh*Ww, 2
+            relative_coords[:, :, 0] += window_size[0] - 1  # shift to start from 0
+            relative_coords[:, :, 1] += window_size[1] - 1
+            relative_coords[:, :, 0] *= 2 * window_size[1] - 1
+            relative_position_index = torch.zeros(
+                size=(window_size[0] * window_size[1] + 1,) * 2, dtype=relative_coords.dtype
+            )
+            relative_position_index[1:, 1:] = relative_coords.sum(-1)  # Wh*Ww, Wh*Ww
+            relative_position_index[0, 0:] = self.num_relative_distance - 3
+            relative_position_index[0:, 0] = self.num_relative_distance - 2
+            relative_position_index[0, 0] = self.num_relative_distance - 1
+
+            self.register_buffer("relative_position_index", relative_position_index)
+        else:
+            self.window_size = None
+            self.relative_position_bias_table = None
+            self.relative_position_index = None
+
+        self.attn_drop = nn.Dropout(attn_drop)
+        self.inner_attn_ln = norm_layer(all_head_dim) if subln else nn.Identity()
+        # self.proj = nn.Linear(all_head_dim, all_head_dim)
+        self.proj = nn.Linear(all_head_dim, dim)
+        self.proj_drop = nn.Dropout(proj_drop)
+        self.xattn = xattn
+        self.xattn_drop = attn_drop
+
+        self.rope = rope
+
+    def forward(self, x, rel_pos_bias=None, attn_mask=None):
+        # B, N, C = x.shape
+
+        B, H, W, C = x.shape
+        x = x.view(B, -1, C)
+        N = H * W
+
+        if self.subln:
+            q = F.linear(input=x, weight=self.q_proj.weight, bias=self.q_bias)
+            k = F.linear(input=x, weight=self.k_proj.weight, bias=None)
+            v = F.linear(input=x, weight=self.v_proj.weight, bias=self.v_bias)
+
+            q = q.reshape(B, N, self.num_heads, -1).permute(0, 2, 1, 3)  # B, num_heads, N, C
+            k = k.reshape(B, N, self.num_heads, -1).permute(0, 2, 1, 3)
+            v = v.reshape(B, N, self.num_heads, -1).permute(0, 2, 1, 3)
+        else:
+
+            qkv_bias = None
+            if self.q_bias is not None:
+                qkv_bias = torch.cat(
+                    (self.q_bias, torch.zeros_like(self.v_bias, requires_grad=False), self.v_bias)
+                )
+
+            qkv = F.linear(input=x, weight=self.qkv.weight, bias=qkv_bias)
+            qkv = qkv.reshape(B, N, 3, self.num_heads, -1).permute(
+                2, 0, 3, 1, 4
+            )  # 3, B, num_heads, N, C
+            q, k, v = qkv[0], qkv[1], qkv[2]
+
+        if self.rope:
+            # slightly fast impl
+            # q_t = q[:, :, 1:, :]
+            # ro_q_t = self.rope(q_t)
+            # q = torch.cat((q[:, :, :1, :], ro_q_t), -2).type_as(v)
+
+            # k_t = k[:, :, 1:, :]
+            # ro_k_t = self.rope(k_t)
+            # k = torch.cat((k[:, :, :1, :], ro_k_t), -2).type_as(v)
+
+            ## rope
+            q = self.rope(q).type_as(v)
+            k = self.rope(k).type_as(v)
+
+        if self.xattn and not (torch.jit.is_scripting() or torch.jit.is_tracing()):
+            q = q.permute(0, 2, 1, 3)  # B, num_heads, N, C -> B, N, num_heads, C
+            k = k.permute(0, 2, 1, 3)
+            v = v.permute(0, 2, 1, 3)
+
+            x = xops.memory_efficient_attention(
+                q,
+                k,
+                v,
+                p=self.xattn_drop,
+                scale=self.scale,
+            )
+            x = x.reshape(B, N, -1)
+            x = self.inner_attn_ln(x)
+            x = self.proj(x)
+            x = self.proj_drop(x)
+        else:
+            q = q * self.scale
+            attn = q @ k.transpose(-2, -1)
+
+            if self.relative_position_bias_table is not None:
+                relative_position_bias = self.relative_position_bias_table[
+                    self.relative_position_index.view(-1)
+                ].view(
+                    self.window_size[0] * self.window_size[1] + 1,
+                    self.window_size[0] * self.window_size[1] + 1,
+                    -1,
+                )  # Wh*Ww,Wh*Ww,nH
+                relative_position_bias = relative_position_bias.permute(
+                    2, 0, 1
+                ).contiguous()  # nH, Wh*Ww, Wh*Ww
+                attn = attn + relative_position_bias.unsqueeze(0).type_as(attn)
+
+            if rel_pos_bias is not None:
+                attn = attn + rel_pos_bias.type_as(attn)
+
+            if attn_mask is not None:
+                attn_mask = attn_mask.bool()
+                attn = attn.masked_fill(~attn_mask[:, None, None, :], float("-inf"))
+
+            attn = attn.softmax(dim=-1)
+            attn = self.attn_drop(attn)
+
+            x = (attn @ v).transpose(1, 2).reshape(B, N, -1)
+            x = self.inner_attn_ln(x)
+            x = self.proj(x)
+            x = self.proj_drop(x)
+
+        x = x.view(B, H, W, C)
+
+        return x
+
+
+class ResBottleneckBlock(CNNBlockBase):
+    """
+    The standard bottleneck residual block without the last activation layer.
+    It contains 3 conv layers with kernels 1x1, 3x3, 1x1.
+    """
+
+    def __init__(
+        self,
+        in_channels,
+        out_channels,
+        bottleneck_channels,
+        norm="LN",
+        act_layer=nn.GELU,
+    ):
+        """
+        Args:
+            in_channels (int): Number of input channels.
+            out_channels (int): Number of output channels.
+            bottleneck_channels (int): number of output channels for the 3x3
+                "bottleneck" conv layers.
+            norm (str or callable): normalization for all conv layers.
+                See :func:`layers.get_norm` for supported format.
+            act_layer (callable): activation for all conv layers.
+        """
+        super().__init__(in_channels, out_channels, 1)
+
+        self.conv1 = Conv2d(in_channels, bottleneck_channels, 1, bias=False)
+        self.norm1 = get_norm(norm, bottleneck_channels)
+        self.act1 = act_layer()
+
+        self.conv2 = Conv2d(
+            bottleneck_channels,
+            bottleneck_channels,
+            3,
+            padding=1,
+            bias=False,
+        )
+        self.norm2 = get_norm(norm, bottleneck_channels)
+        self.act2 = act_layer()
+
+        self.conv3 = Conv2d(bottleneck_channels, out_channels, 1, bias=False)
+        self.norm3 = get_norm(norm, out_channels)
+
+        for layer in [self.conv1, self.conv2, self.conv3]:
+            weight_init.c2_msra_fill(layer)
+        for layer in [self.norm1, self.norm2]:
+            layer.weight.data.fill_(1.0)
+            layer.bias.data.zero_()
+        # zero init last norm layer.
+        self.norm3.weight.data.zero_()
+        self.norm3.bias.data.zero_()
+
+    def forward(self, x):
+        out = x
+        for layer in self.children():
+            out = layer(out)
+
+        out = x + out
+        return out
+
+
+class Block(nn.Module):
+    """Transformer blocks with support of window attention and residual propagation blocks"""
+
+    def __init__(
+        self,
+        dim,
+        num_heads,
+        mlp_ratio=4.0,
+        qkv_bias=False,
+        qk_scale=None,
+        drop=0.0,
+        attn_drop=0.0,
+        drop_path=0.0,
+        init_values=None,
+        act_layer=nn.GELU,
+        norm_layer=partial(nn.LayerNorm, eps=1e-6), 
+        window_size=0,
+        use_residual_block=False,
+        attn_head_dim=None,
+        rope=None,
+        xattn=False,
+        postnorm=False,
+        subln=False,
+        naiveswiglu=False,
+    ):
+        """
+        Args:
+            dim (int): Number of input channels.
+            num_heads (int): Number of attention heads in each ViT block.
+            mlp_ratio (float): Ratio of mlp hidden dim to embedding dim.
+            qkv_bias (bool): If True, add a learnable bias to query, key, value.
+            drop_path (float): Stochastic depth rate.
+            norm_layer (nn.Module): Normalization layer.
+            act_layer (nn.Module): Activation layer.
+            use_rel_pos (bool): If True, add relative positional embeddings to the attention map.
+            rel_pos_zero_init (bool): If True, zero initialize relative positional parameters.
+            window_size (int): Window size for window attention blocks. If it equals 0, then not
+                use window attention.
+            use_residual_block (bool): If True, use a residual block after the MLP block.
+            input_size (int or None): Input resolution for calculating the relative positional
+                parameter size.
+        """
+        super().__init__()
+        self.norm1 = norm_layer(dim)
+        self.attn = Attention(
+            dim,
+            num_heads=num_heads,
+            qkv_bias=qkv_bias,
+            qk_scale=qk_scale,
+            attn_drop=attn_drop,
+            proj_drop=drop,
+            window_size=window_size,
+            attn_head_dim=attn_head_dim,
+            rope=rope,
+            xattn=xattn,
+            subln=subln,
+            norm_layer=norm_layer,
+        )
+
+        from timm.models.layers import DropPath
+
+        self.drop_path = DropPath(drop_path) if drop_path > 0.0 else nn.Identity()
+        self.norm2 = norm_layer(dim)
+        mlp_hidden_dim = int(dim * mlp_ratio)
+
+        if naiveswiglu:
+            self.mlp = SwiGLU(
+                in_features=dim,
+                hidden_features=mlp_hidden_dim,
+                subln=subln,
+                norm_layer=norm_layer,
+            )
+        else:
+            self.mlp = Mlp(
+                in_features=dim,
+                hidden_features=mlp_hidden_dim,
+                act_layer=act_layer,
+                subln=subln,
+                drop=drop,
+            )
+
+        if init_values is not None and init_values > 0:
+            self.gamma_1 = nn.Parameter(init_values * torch.ones((dim)), requires_grad=True)
+            self.gamma_2 = nn.Parameter(init_values * torch.ones((dim)), requires_grad=True)
+        else:
+            self.gamma_1, self.gamma_2 = None, None
+
+        self.postnorm = postnorm
+
+        self.window_size = window_size
+
+        self.use_residual_block = use_residual_block
+        if use_residual_block:
+            # Use a residual block with bottleneck channel as dim // 2
+            self.residual = ResBottleneckBlock(
+                in_channels=dim,
+                out_channels=dim,
+                bottleneck_channels=dim // 2,
+                norm="LN",
+            )
+
+    def forward(self, x, rel_pos_bias=None, attn_mask=None):
+        if self.gamma_1 is None:
+            if self.postnorm:
+                shortcut = x
+
+                # Window partition
+                if self.window_size > 0:
+                    H, W = x.shape[1], x.shape[2]
+                    x, pad_hw = window_partition(x, self.window_size)
+
+                x = self.attn(x, rel_pos_bias=rel_pos_bias, attn_mask=attn_mask)
+
+                # Reverse window partition
+                if self.window_size > 0:
+                    x = window_unpartition(x, self.window_size, pad_hw, (H, W))
+
+                x = self.norm1(x)
+                x = shortcut + self.drop_path(x)
+
+                # x = x + self.drop_path(self.norm1(self.attn(x, rel_pos_bias=rel_pos_bias, attn_mask=attn_mask)))
+                x = x + self.drop_path(self.norm2(self.mlp(x)))
+            else:
+                shortcut = x
+                x = self.norm1(x)
+
+                # Window partition
+                if self.window_size > 0:
+                    H, W = x.shape[1], x.shape[2]
+                    x, pad_hw = window_partition(x, self.window_size)
+
+                x = self.attn(x, rel_pos_bias=rel_pos_bias, attn_mask=attn_mask)
+
+                # Reverse window partition
+                if self.window_size > 0:
+                    x = window_unpartition(x, self.window_size, pad_hw, (H, W))
+
+                x = shortcut + self.drop_path(x)
+
+                # x = x + self.drop_path(self.attn(self.norm1(x), rel_pos_bias=rel_pos_bias, attn_mask=attn_mask))
+                x = x + self.drop_path(self.mlp(self.norm2(x)))
+        else:
+            if self.postnorm:
+                shortcut = x
+
+                # Window partition
+                if self.window_size > 0:
+                    H, W = x.shape[1], x.shape[2]
+                    x, pad_hw = window_partition(x, self.window_size)
+
+                x = self.attn(x, rel_pos_bias=rel_pos_bias, attn_mask=attn_mask)
+
+                # Reverse window partition
+                if self.window_size > 0:
+                    x = window_unpartition(x, self.window_size, pad_hw, (H, W))
+
+                x = self.norm1(x)
+                x = shortcut + self.drop_path(self.gamma_1 * x)
+
+                # x = x + self.drop_path(self.gamma_1 * self.norm1(self.attn(x, rel_pos_bias=rel_pos_bias, attn_mask=attn_mask)))
+                x = x + self.drop_path(self.gamma_2 * self.norm2(self.mlp(x)))
+            else:
+                shortcut = x
+                x = self.norm1(x)
+
+                # Window partition
+                if self.window_size > 0:
+                    H, W = x.shape[1], x.shape[2]
+                    x, pad_hw = window_partition(x, self.window_size)
+
+                x = self.attn(x, rel_pos_bias=rel_pos_bias, attn_mask=attn_mask)
+
+                # Reverse window partition
+                if self.window_size > 0:
+                    x = window_unpartition(x, self.window_size, pad_hw, (H, W))
+
+                x = shortcut + self.drop_path(self.gamma_1 * x)
+
+                # x = x + self.drop_path(self.gamma_1 * self.attn(self.norm1(x), rel_pos_bias=rel_pos_bias, attn_mask=attn_mask))
+                x = x + self.drop_path(self.gamma_2 * self.mlp(self.norm2(x)))
+
+        if self.use_residual_block:
+            x = self.residual(x.permute(0, 3, 1, 2)).permute(0, 2, 3, 1)
+
+        return x
+
+
+class ViT(Backbone):
+    """
+    This module implements Vision Transformer (ViT) backbone in :paper:`vitdet`.
+    "Exploring Plain Vision Transformer Backbones for Object Detection",
+    https://arxiv.org/abs/2203.16527
+    """
+
+    def __init__(
+        self,
+        img_size=1024,
+        patch_size=16,
+        in_chans=3,
+        embed_dim=768,
+        depth=12,
+        num_heads=12,
+        mlp_ratio=4.0,
+        qkv_bias=False,
+        qk_scale=None,
+        drop_rate=0.0,
+        attn_drop_rate=0.0,
+        drop_path_rate=0.0,
+        norm_layer=partial(nn.LayerNorm, eps=1e-6),
+        init_values=None,
+        use_abs_pos=True,
+        use_rel_pos=False,
+        rope=False,
+        postnorm=False,
+        pt_hw_seq_len=16,
+        intp_freq=False,
+        naiveswiglu=False,
+        subln=False,
+        window_size=0,
+        window_block_indexes=(),
+        residual_block_indexes=(),
+        use_act_checkpoint=False,
+        pretrain_img_size=224,
+        pretrain_use_cls_token=True,
+        out_feature="last_feat",
+        xattn=False,
+        frozen_stages=-1,
+    ):
+        """
+        Args:
+            img_size (int): Input image size.
+            patch_size (int): Patch size.
+            in_chans (int): Number of input image channels.
+            embed_dim (int): Patch embedding dimension.
+            depth (int): Depth of ViT.
+            num_heads (int): Number of attention heads in each ViT block.
+            mlp_ratio (float): Ratio of mlp hidden dim to embedding dim.
+            qkv_bias (bool): If True, add a learnable bias to query, key, value.
+            drop_path_rate (float): Stochastic depth rate.
+            norm_layer (nn.Module): Normalization layer.
+            act_layer (nn.Module): Activation layer.
+            use_abs_pos (bool): If True, use absolute positional embeddings.
+            use_rel_pos (bool): If True, add relative positional embeddings to the attention map.
+            rel_pos_zero_init (bool): If True, zero initialize relative positional parameters.
+            window_size (int): Window size for window attention blocks.
+            window_block_indexes (list): Indexes for blocks using window attention.
+            residual_block_indexes (list): Indexes for blocks using conv propagation.
+            use_act_checkpoint (bool): If True, use activation checkpointing.
+            pretrain_img_size (int): input image size for pretraining models.
+            pretrain_use_cls_token (bool): If True, pretrainig models use class token.
+            out_feature (str): name of the feature from the last block.
+        """
+        super().__init__()
+        self.pretrain_use_cls_token = pretrain_use_cls_token
+
+        self.patch_embed = PatchEmbed(
+            kernel_size=(patch_size, patch_size),
+            stride=(patch_size, patch_size),
+            in_chans=in_chans,
+            embed_dim=embed_dim,
+        )
+
+        if use_abs_pos:
+            # Initialize absolute positional embedding with pretrain image size.
+            num_patches = (pretrain_img_size // patch_size) * (pretrain_img_size // patch_size)
+            num_positions = (num_patches + 1) if pretrain_use_cls_token else num_patches
+            self.pos_embed = nn.Parameter(torch.zeros(1, num_positions, embed_dim))
+        else:
+            self.pos_embed = None
+
+        if rope:
+            half_head_dim = embed_dim // num_heads // 2
+            hw_seq_len = img_size // patch_size
+            self.rope_win = VisionRotaryEmbeddingFast(
+                dim=half_head_dim,
+                pt_seq_len=pt_hw_seq_len,
+                ft_seq_len=window_size if intp_freq else None,
+            )
+            self.rope_glb = VisionRotaryEmbeddingFast(
+                dim=half_head_dim,
+                pt_seq_len=pt_hw_seq_len,
+                ft_seq_len=hw_seq_len if intp_freq else None,
+            )
+        else:
+            self.rope_win = None
+            self.rope_glb = None
+
+        self.naiveswiglu = naiveswiglu
+
+        # stochastic depth decay rule
+        dpr = [x.item() for x in torch.linspace(0, drop_path_rate, depth)]
+
+        self.blocks = nn.ModuleList()
+        for i in range(depth):
+            block = Block(
+                dim=embed_dim,
+                num_heads=num_heads,
+                mlp_ratio=mlp_ratio,
+                qkv_bias=qkv_bias,
+                qk_scale=qk_scale,
+                drop=drop_rate,
+                attn_drop=attn_drop_rate,
+                drop_path=dpr[i],
+                norm_layer=norm_layer,
+                init_values=init_values,
+                window_size=window_size if i in window_block_indexes else 0,
+                postnorm=postnorm,
+                subln=subln,
+                naiveswiglu=naiveswiglu,
+                use_residual_block=i in residual_block_indexes,
+                rope=self.rope_win if i in window_block_indexes else self.rope_glb,
+                xattn=xattn,
+            )
+            if use_act_checkpoint and i > frozen_stages - 1:
+                # TODO: use torch.utils.checkpoint
+                from fairscale.nn.checkpoint import checkpoint_wrapper
+
+                block = checkpoint_wrapper(block)
+            self.blocks.append(block)
+
+        self._out_feature_channels = {out_feature: embed_dim}
+        self._out_feature_strides = {out_feature: patch_size}
+        self._out_features = [out_feature]
+
+        if self.pos_embed is not None:
+            nn.init.trunc_normal_(self.pos_embed, std=0.02)
+
+        self.apply(self._init_weights)
+
+        self.frozen_stages = frozen_stages
+        self._freeze_stages()
+
+    def _freeze_stages(self):
+        if self.frozen_stages >= 0:
+            self.patch_embed.eval()
+            for param in self.patch_embed.parameters():
+                param.requires_grad = False
+
+        if self.frozen_stages >= 1 and self.pos_embed is not None:
+            self.pos_embed.requires_grad = False
+
+        if self.frozen_stages >= 2:
+            for i in range(0, self.frozen_stages - 1):
+                m = self.blocks[i]
+                m.eval()
+                for name, param in m.named_parameters():
+                    vit_lr_decay_rate = get_vit_lr_decay_rate(f"backbone.net.blocks.{i}.{name}", lr_decay_rate=0.9, num_layers=len(self.blocks))
+                    logger.info(f"freeze blocks.{i}.{name} {param.size()} {vit_lr_decay_rate}")
+                    param.requires_grad = False
+
+
+    def _init_weights(self, m):
+        if isinstance(m, nn.Linear):
+            nn.init.trunc_normal_(m.weight, std=0.02)
+            if isinstance(m, nn.Linear) and m.bias is not None:
+                nn.init.constant_(m.bias, 0)
+        elif isinstance(m, nn.LayerNorm):
+            nn.init.constant_(m.bias, 0)
+            nn.init.constant_(m.weight, 1.0)
+
+    def forward(self, x):
+        x = self.patch_embed(x)
+        if self.pos_embed is not None:
+            x = x + get_abs_pos(
+                self.pos_embed, self.pretrain_use_cls_token, (x.shape[1], x.shape[2])
+            )
+
+        for blk in self.blocks:
+            x = blk(x)
+
+        outputs = {self._out_features[0]: x.permute(0, 3, 1, 2)}
+        return outputs
+
+
+class SimpleFeaturePyramid(Backbone):
+    """
+    This module implements SimpleFeaturePyramid in :paper:`vitdet`.
+    It creates pyramid features built on top of the input feature map.
+    """
+
+    def __init__(
+        self,
+        net,
+        in_feature,
+        out_channels,
+        scale_factors,
+        top_block=None,
+        norm="LN",
+        square_pad=0,
+    ):
+        """
+        Args:
+            net (Backbone): module representing the subnetwork backbone.
+                Must be a subclass of :class:`Backbone`.
+            in_feature (str): names of the input feature maps coming
+                from the net.
+            out_channels (int): number of channels in the output feature maps.
+            scale_factors (list[float]): list of scaling factors to upsample or downsample
+                the input features for creating pyramid features.
+            top_block (nn.Module or None): if provided, an extra operation will
+                be performed on the output of the last (smallest resolution)
+                pyramid output, and the result will extend the result list. The top_block
+                further downsamples the feature map. It must have an attribute
+                "num_levels", meaning the number of extra pyramid levels added by
+                this block, and "in_feature", which is a string representing
+                its input feature (e.g., p5).
+            norm (str): the normalization to use.
+            square_pad (int): If > 0, require input images to be padded to specific square size.
+        """
+        super(SimpleFeaturePyramid, self).__init__()
+        assert isinstance(net, Backbone)
+
+        self.scale_factors = scale_factors
+
+        input_shapes = net.output_shape()
+        strides = [int(input_shapes[in_feature].stride / scale) for scale in scale_factors]
+        _assert_strides_are_log2_contiguous(strides)
+
+        dim = input_shapes[in_feature].channels
+        self.stages = []
+        use_bias = norm == ""
+        for idx, scale in enumerate(scale_factors):
+            out_dim = dim
+            if scale == 4.0:
+                layers = [
+                    nn.ConvTranspose2d(dim, dim // 2, kernel_size=2, stride=2),
+                    get_norm(norm, dim // 2),
+                    nn.GELU(),
+                    nn.ConvTranspose2d(dim // 2, dim // 4, kernel_size=2, stride=2),
+                ]
+                out_dim = dim // 4
+            elif scale == 2.0:
+                layers = [nn.ConvTranspose2d(dim, dim // 2, kernel_size=2, stride=2)]
+                out_dim = dim // 2
+            elif scale == 1.0:
+                layers = []
+            elif scale == 0.5:
+                layers = [nn.MaxPool2d(kernel_size=2, stride=2)]
+            else:
+                raise NotImplementedError(f"scale_factor={scale} is not supported yet.")
+
+            layers.extend(
+                [
+                    Conv2d(
+                        out_dim,
+                        out_channels,
+                        kernel_size=1,
+                        bias=use_bias,
+                        norm=get_norm(norm, out_channels),
+                    ),
+                    Conv2d(
+                        out_channels,
+                        out_channels,
+                        kernel_size=3,
+                        padding=1,
+                        bias=use_bias,
+                        norm=get_norm(norm, out_channels),
+                    ),
+                ]
+            )
+            layers = nn.Sequential(*layers)
+
+            stage = int(math.log2(strides[idx]))
+            self.add_module(f"simfp_{stage}", layers)
+            self.stages.append(layers)
+
+        self.net = net
+        self.in_feature = in_feature
+        self.top_block = top_block
+        # Return feature names are "p<stage>", like ["p2", "p3", ..., "p6"]
+        self._out_feature_strides = {"p{}".format(int(math.log2(s))): s for s in strides}
+        # top block output feature maps.
+        if self.top_block is not None:
+            for s in range(stage, stage + self.top_block.num_levels):
+                self._out_feature_strides["p{}".format(s + 1)] = 2 ** (s + 1)
+
+        self._out_features = list(self._out_feature_strides.keys())
+        self._out_feature_channels = {k: out_channels for k in self._out_features}
+        self._size_divisibility = strides[-1]
+        self._square_pad = square_pad
+
+    @property
+    def padding_constraints(self):
+        return {
+            "size_divisiblity": self._size_divisibility,
+            "square_size": self._square_pad,
+        }
+
+    def forward(self, x):
+        """
+        Args:
+            x: Tensor of shape (N,C,H,W). H, W must be a multiple of ``self.size_divisibility``.
+
+        Returns:
+            dict[str->Tensor]:
+                mapping from feature map name to pyramid feature map tensor
+                in high to low resolution order. Returned feature names follow the FPN
+                convention: "p<stage>", where stage has stride = 2 ** stage e.g.,
+                ["p2", "p3", ..., "p6"].
+        """
+        bottom_up_features = self.net(x)
+        features = bottom_up_features[self.in_feature]
+        results = []
+
+        for stage in self.stages:
+            results.append(stage(features))
+            # with torch.cuda.amp.autocast(enabled=False):
+            #     results.append(stage(features.float()))
+            if torch.any(torch.isnan(results[-1])):
+                v = results[-1]
+                print("stage", len(results), v, v.size(), v.max(), v.min(), "all finite:", torch.all(torch.isfinite(v)), "any nan:", torch.any(torch.isnan(v)))
+                v = features
+                print(self.in_feature, v, v.size(), v.max(), v.min(), "all finite:", torch.all(torch.isfinite(v)), "any nan:", torch.any(torch.isnan(v)))
+                print("stage parameters", stage, sum([_.sum() for _ in stage.parameters()]))
+                for name, param in stage.named_parameters():
+                    print(name, param.size(), param.sum(), param.max(), param.min(), "all finite:", torch.all(torch.isfinite(param)), "any nan:", torch.any(torch.isnan(param)))
+                for name, buf in stage.named_buffers():
+                    print(name, buf.size(), buf.sum(), buf.max(), buf.min(), "all finite:", torch.all(torch.isfinite(buf)), "any nan:", torch.any(torch.isnan(buf)))
+
+                with torch.cuda.amp.autocast(enabled=False):
+                    results.pop()
+                    results.append(stage(features.float()))
+                    # results[-1] = stage(features.float())
+
+        if self.top_block is not None:
+            if self.top_block.in_feature in bottom_up_features:
+                top_block_in_feature = bottom_up_features[self.top_block.in_feature]
+            else:
+                top_block_in_feature = results[self._out_features.index(self.top_block.in_feature)]
+            results.extend(self.top_block(top_block_in_feature))
+            # with torch.cuda.amp.autocast(enabled=False):
+            #     results.extend(self.top_block(top_block_in_feature.float()))
+            if torch.any(torch.isnan(results[-1])):
+                v = results[-1]
+                print(len(results), v, v.size(), v.max(), v.min(), "all finite:", torch.all(torch.isfinite(v)), "any nan:", torch.any(torch.isnan(v)))
+                v = top_block_in_feature
+                print(self.top_block.in_feature, v, v.size(), v.max(), v.min(), "all finite:", torch.all(torch.isfinite(v)), "any nan:", torch.any(torch.isnan(v)))
+                print("top_block parameters", self.top_block, sum([_.sum() for _ in self.top_block.parameters()]))
+        assert len(self._out_features) == len(results)
+        return {f: res for f, res in zip(self._out_features, results)}
+
+
+def get_vit_lr_decay_rate(name, lr_decay_rate=1.0, num_layers=12):
+    """
+    Calculate lr decay rate for different ViT blocks.
+    Args:
+        name (string): parameter name.
+        lr_decay_rate (float): base lr decay rate.
+        num_layers (int): number of ViT blocks.
+
+    Returns:
+        lr decay rate for the given parameter.
+    """
+    if name.startswith("model_vision."):
+        name = name[len("model_vision.") :]
+
+    layer_id = num_layers + 1
+    if name.startswith("backbone"):
+        if ".pos_embed" in name or ".patch_embed" in name:
+            layer_id = 0
+        elif ".blocks." in name and ".residual." not in name:
+            layer_id = int(name[name.find(".blocks.") :].split(".")[2]) + 1
+
+    logger.info(
+        "get_vit_lr_decay_rate: name={} num_layers={} layer_id={} lr_decay_rate={}".format(
+            name, num_layers, layer_id, lr_decay_rate ** (num_layers + 1 - layer_id)
+        )
+    )
+    return lr_decay_rate ** (num_layers + 1 - layer_id)

approach/ovod/APE/ape/modeling/text/bert_wrapper.py

@@ -0,0 +1,107 @@
+import torch
+from torch import nn
+from torch.cuda.amp import autocast
+
+from transformers import (
+    AutoConfig,
+    AutoModelForSeq2SeqLM,
+    AutoTokenizer,
+    BertConfig,
+    BertModel,
+    RobertaConfig,
+    RobertaModel,
+)
+
+
+class Bert(nn.Module):
+    def __init__(
+        self,
+        pretrained_model_name_or_path,
+        dtype="float32",
+        **kwargs,
+    ):
+        super().__init__(**kwargs)
+
+        self.dtype = getattr(torch, dtype)
+
+        self.config = BertConfig.from_pretrained(
+            pretrained_model_name_or_path=pretrained_model_name_or_path
+        )
+        self.bert_model = BertModel.from_pretrained(
+            pretrained_model_name_or_path=pretrained_model_name_or_path,
+            add_pooling_layer=False,
+        )
+        self.tokenizer = AutoTokenizer.from_pretrained(
+            pretrained_model_name_or_path=pretrained_model_name_or_path
+        )
+
+        self.bert_model.eval()
+        for name, param in self.bert_model.named_parameters():
+            param.requires_grad = False
+            param.data = param.data.to(self.dtype)
+
+        self.register_buffer("unused_tensor", torch.zeros(1), False)
+
+        self.text_list_to_feature = {}
+
+    @property
+    def device(self):
+        return self.unused_tensor.device
+
+    @autocast(enabled=False)
+    @torch.no_grad()
+    def forward_text(self, text_list, cache=False):
+
+        if cache and tuple(text_list) in self.text_list_to_feature:
+            return self.text_list_to_feature[tuple(text_list)]
+
+        tokenized = self.tokenizer.batch_encode_plus(
+            text_list,
+            max_length=256,
+            padding="max_length" if True else "longest",
+            return_special_tokens_mask=True,
+            return_tensors="pt",
+            truncation=True,
+        ).to(self.device)
+
+        input_ids = tokenized.input_ids  # (bs, seq_len)
+        attention_mask = tokenized.attention_mask  # (bs, seq_len)
+
+        max_batch_size = 500
+        if len(input_ids) > max_batch_size:
+            chunck_num = len(input_ids) // max_batch_size + 1
+            outputss = [
+                self.bert_model(
+                    input_ids=input_ids[
+                        chunck_id * max_batch_size : (chunck_id + 1) * max_batch_size
+                    ],
+                    attention_mask=attention_mask[
+                        chunck_id * max_batch_size : (chunck_id + 1) * max_batch_size
+                    ],
+                )
+                for chunck_id in range(chunck_num)
+            ]
+
+            last_hidden_state = torch.cat(
+                [outputs.last_hidden_state for outputs in outputss], dim=0
+            )
+        else:
+            outputs = self.bert_model(
+                input_ids=input_ids,
+                attention_mask=attention_mask,
+            )
+
+            last_hidden_state = outputs.last_hidden_state
+
+        end_token_idx = input_ids.argmin(dim=-1) - 1
+
+        ret = {
+            "end_token_idx": end_token_idx,
+            "attention_mask": attention_mask,
+            "last_hidden_state": last_hidden_state,
+        }
+
+        if cache:
+            self.text_list_to_feature[tuple(text_list)] = ret
+
+        return ret

approach/ovod/APE/ape/modeling/text/clip_wrapper.py

@@ -0,0 +1,224 @@
+import logging
+from collections import OrderedDict
+from typing import List, Union
+
+import torch
+from torch import nn
+
+from clip.simple_tokenizer import SimpleTokenizer as _Tokenizer
+
+__all__ = ["tokenize"]
+
+count = 0
+
+
+class LayerNorm(nn.LayerNorm):
+    """Subclass torch's LayerNorm to handle fp16."""
+
+    def forward(self, x: torch.Tensor):
+        orig_type = x.dtype
+        ret = super().forward(x.type(torch.float32))
+        return ret.type(orig_type)
+
+
+class QuickGELU(nn.Module):
+    def forward(self, x: torch.Tensor):
+        return x * torch.sigmoid(1.702 * x)
+
+
+class ResidualAttentionBlock(nn.Module):
+    def __init__(self, d_model: int, n_head: int, attn_mask: torch.Tensor = None):
+        super().__init__()
+
+        self.attn = nn.MultiheadAttention(d_model, n_head)
+        self.ln_1 = LayerNorm(d_model)
+        self.mlp = nn.Sequential(
+            OrderedDict(
+                [
+                    ("c_fc", nn.Linear(d_model, d_model * 4)),
+                    ("gelu", QuickGELU()),
+                    ("c_proj", nn.Linear(d_model * 4, d_model)),
+                ]
+            )
+        )
+        self.ln_2 = LayerNorm(d_model)
+        self.attn_mask = attn_mask
+
+    def attention(self, x: torch.Tensor):
+        self.attn_mask = (
+            self.attn_mask.to(dtype=x.dtype, device=x.device)
+            if self.attn_mask is not None
+            else None
+        )
+        return self.attn(x, x, x, need_weights=False, attn_mask=self.attn_mask)[0]
+
+    def forward(self, x: torch.Tensor):
+        x = x + self.attention(self.ln_1(x))
+        x = x + self.mlp(self.ln_2(x))
+        return x
+
+
+class Transformer(nn.Module):
+    def __init__(self, width: int, layers: int, heads: int, attn_mask: torch.Tensor = None):
+        super().__init__()
+        self.width = width
+        self.layers = layers
+        self.resblocks = nn.Sequential(
+            *[ResidualAttentionBlock(width, heads, attn_mask) for _ in range(layers)]
+        )
+
+    def forward(self, x: torch.Tensor):
+        return self.resblocks(x)
+
+
+class CLIPTEXT(nn.Module):
+    def __init__(
+        self,
+        embed_dim=512,
+        context_length=77,
+        vocab_size=49408,
+        transformer_width=512,
+        transformer_heads=8,
+        transformer_layers=12,
+    ):
+        super().__init__()
+
+        self._tokenizer = _Tokenizer()
+        self.context_length = context_length
+
+        self.transformer = Transformer(
+            width=transformer_width,
+            layers=transformer_layers,
+            heads=transformer_heads,
+            attn_mask=self.build_attention_mask(),
+        )
+
+        self.vocab_size = vocab_size
+        self.token_embedding = nn.Embedding(vocab_size, transformer_width)
+        self.positional_embedding = nn.Parameter(
+            torch.empty(self.context_length, transformer_width)
+        )
+        self.ln_final = LayerNorm(transformer_width)
+
+        self.text_projection = nn.Parameter(torch.empty(transformer_width, embed_dim))
+
+        self.initialize_parameters()
+
+    def initialize_parameters(self):
+        nn.init.normal_(self.token_embedding.weight, std=0.02)
+        nn.init.normal_(self.positional_embedding, std=0.01)
+
+        proj_std = (self.transformer.width**-0.5) * ((2 * self.transformer.layers) ** -0.5)
+        attn_std = self.transformer.width**-0.5
+        fc_std = (2 * self.transformer.width) ** -0.5
+        for block in self.transformer.resblocks:
+            nn.init.normal_(block.attn.in_proj_weight, std=attn_std)
+            nn.init.normal_(block.attn.out_proj.weight, std=proj_std)
+            nn.init.normal_(block.mlp.c_fc.weight, std=fc_std)
+            nn.init.normal_(block.mlp.c_proj.weight, std=proj_std)
+
+        if self.text_projection is not None:
+            nn.init.normal_(self.text_projection, std=self.transformer.width**-0.5)
+
+    def build_attention_mask(self):
+        mask = torch.empty(self.context_length, self.context_length)
+        mask.fill_(float("-inf"))
+        mask.triu_(1)  # zero out the lower diagonal
+        return mask
+
+    @property
+    def device(self):
+        return self.text_projection.device
+
+    @property
+    def dtype(self):
+        return self.text_projection.dtype
+
+    def tokenize(self, texts: Union[str, List[str]], context_length: int = 77) -> torch.LongTensor:
+        """ """
+        if isinstance(texts, str):
+            texts = [texts]
+
+        sot_token = self._tokenizer.encoder["<|startoftext|>"]
+        eot_token = self._tokenizer.encoder["<|endoftext|>"]
+        all_tokens = [[sot_token] + self._tokenizer.encode(text) + [eot_token] for text in texts]
+        result = torch.zeros(len(all_tokens), context_length, dtype=torch.long)
+
+        for i, tokens in enumerate(all_tokens):
+            if len(tokens) > context_length:
+                st = torch.randint(len(tokens) - context_length + 1, (1,))[0].item()
+                tokens = tokens[st : st + context_length]
+            result[i, : len(tokens)] = torch.tensor(tokens)
+
+        return result
+
+    def encode_text(self, text):
+        x = self.token_embedding(text).type(self.dtype)  # [batch_size, n_ctx, d_model]
+        x = x + self.positional_embedding.type(self.dtype)
+        x = x.permute(1, 0, 2)  # NLD -> LND
+        x = self.transformer(x)
+        x = x.permute(1, 0, 2)  # LND -> NLD
+        x = self.ln_final(x).type(self.dtype)
+        x = x[torch.arange(x.shape[0]), text.argmax(dim=-1)] @ self.text_projection
+        return x
+
+    def forward(self, captions):
+        """
+        captions: list of strings
+        """
+        text = self.tokenize(captions).to(self.device)  # B x L x D
+        features = self.encode_text(text)  # B x D
+        return features
+
+
+def build_clip_text_encoder(model_path, pretrain=True):
+    logger = logging.getLogger(__name__)
+    if pretrain:
+        logger.info("Loading pretrained CLIP " + model_path)
+        import clip
+
+        print(model_path)
+        pretrained_model, _ = clip.load(model_path, device="cpu")
+        state_dict = pretrained_model.state_dict()
+        to_delete_keys = ["logit_scale", "input_resolution", "context_length", "vocab_size"] + [
+            k for k in state_dict.keys() if k.startswith("visual.")
+        ]
+        for k in to_delete_keys:
+            if k in state_dict:
+                del state_dict[k]
+
+        embed_dim = state_dict["text_projection"].shape[1]
+        context_length = state_dict["positional_embedding"].shape[0]
+        vocab_size = state_dict["token_embedding.weight"].shape[0]
+        transformer_width = state_dict["ln_final.weight"].shape[0]
+        transformer_heads = transformer_width // 64
+        transformer_layers = len(
+            set(k.split(".")[2] for k in state_dict if k.startswith(f"transformer.resblocks"))
+        )
+
+        text_encoder = CLIPTEXT(
+            embed_dim,
+            context_length,
+            vocab_size,
+            transformer_width,
+            transformer_heads,
+            transformer_layers,
+        )
+        text_encoder.load_state_dict(state_dict)
+
+    else:
+        logger.info("Building CLIPTEXT")
+        text_encoder = CLIPTEXT(embed_dim=embed_dim)
+    return text_encoder
+
+
+def get_clip_embeddings(text_model, vocabulary, prompt="a "):
+    if isinstance(text_model, str):
+        text_encoder = build_clip_text_encoder(text_model, pretrain=True)
+        text_encoder.eval()
+    else:
+        text_encoder = text_model
+    text_encoder.eval()
+    texts = [prompt + x for x in vocabulary]
+    emb = text_encoder(texts).detach().contiguous()
+    return emb

approach/ovod/APE/ape/modeling/text/clip_wrapper_eva02.py

@@ -0,0 +1,148 @@
+import torch
+import torch.nn as nn
+from torch.cuda.amp import autocast
+
+from .eva02_clip import create_model_and_transforms, get_tokenizer
+
+
+class EVA02CLIP(nn.Module):
+    def __init__(
+        self,
+        clip_model="EVA02-CLIP-B-16",
+        cache_dir="EVA02_CLIP_B_psz16_s8B.pt",
+        dtype="float32",
+        max_batch_size=2560,
+    ):
+        super().__init__()
+        self.net, _, _ = create_model_and_transforms(
+            clip_model, pretrained=cache_dir, force_custom_clip=True
+        )
+        self.tokenizer = get_tokenizer(clip_model)
+
+        if dtype == "bfloat16":
+            self.dtype = torch.bfloat16
+        elif dtype == "float16":
+            self.dtype = torch.float16
+        else:
+            self.dtype = torch.float32
+
+        del self.net.visual
+        self.net.eval()
+        for name, param in self.net.named_parameters():
+            param.requires_grad = False
+            param.data = param.data.to(self.dtype)
+
+        self.register_buffer("unused_tensor", torch.zeros(1), False)
+
+        self.text_list_to_feature = {}
+
+        self.max_batch_size = max_batch_size
+
+    @property
+    def device(self):
+        return self.unused_tensor.device
+
+    def infer_image(self, features):
+        x = features["image"][0]
+        x = self.net.encode_image(x)
+        return x
+
+    @autocast(enabled=False)
+    @torch.no_grad()
+    def encode_text(self, text_list, cache=False):
+        if cache and tuple(text_list) in self.text_list_to_feature:
+            return self.text_list_to_feature[tuple(text_list)]
+
+        text_token = self.tokenizer(text_list, context_length=77).to(self.device)
+
+        max_batch_size = self.max_batch_size
+        if self.device.type == "cpu" or torch.cuda.mem_get_info(self.device)[0] / 1024**3 < 5:
+            max_batch_size = min(256, max_batch_size)
+        if len(text_token) > max_batch_size:
+            chunck_num = len(text_token) // max_batch_size + 1
+            encoder_outputs = torch.cat(
+                [
+                    self.net.encode_text(
+                        text_token[chunck_id * max_batch_size : (chunck_id + 1) * max_batch_size]
+                    )
+                    for chunck_id in range(chunck_num)
+                ],
+                dim=0,
+            )
+        else:
+            encoder_outputs = self.net.encode_text(text_token)
+
+        ret = {
+            "last_hidden_state_eot": encoder_outputs,
+        }
+
+        if cache:
+            self.text_list_to_feature[tuple(text_list)] = ret
+
+        return ret
+
+    @autocast(enabled=False)
+    @torch.no_grad()
+    def forward_text(self, text_list, cache=False):
+        if cache and tuple(text_list) in self.text_list_to_feature:
+            return self.text_list_to_feature[tuple(text_list)]
+
+        text_token = self.tokenizer(text_list, context_length=77).to(self.device)
+
+        max_batch_size = self.max_batch_size
+        if self.device.type == "cpu" or torch.cuda.mem_get_info(self.device)[0] / 1024**3 < 5:
+            max_batch_size = min(256, max_batch_size)
+        if len(text_token) > max_batch_size:
+            chunck_num = len(text_token) // max_batch_size + 1
+            encoder_outputs = [
+                self.custom_encode_text(
+                    text_token[chunck_id * max_batch_size : (chunck_id + 1) * max_batch_size],
+                    self.net.text,
+                )
+                for chunck_id in range(chunck_num)
+            ]
+            encoder_outputs_x = torch.cat([x for (x, _) in encoder_outputs], dim=0)
+            encoder_outputs_xx = torch.cat([xx for (_, xx) in encoder_outputs], dim=0)
+        else:
+            encoder_outputs_x, encoder_outputs_xx = self.custom_encode_text(
+                text_token, self.net.text
+            )
+
+        end_token_idx = text_token.argmax(dim=-1)
+        attention_mask = end_token_idx.new_zeros(encoder_outputs_xx.size()[:2])
+        for i in range(attention_mask.size(0)):
+            attention_mask[i, : end_token_idx[i] + 1] = 1
+
+        ret = {
+            "end_token_idx": end_token_idx,
+            "attention_mask": attention_mask,
+            "last_hidden_state": encoder_outputs_xx,
+            "last_hidden_state_eot": encoder_outputs_x,
+        }
+
+        if cache:
+            self.text_list_to_feature[tuple(text_list)] = ret
+
+        return ret
+
+    @autocast(enabled=False)
+    @torch.no_grad()
+    def custom_encode_text(self, text, m, normalize: bool = False):
+        cast_dtype = m.transformer.get_cast_dtype()
+
+        x = m.token_embedding(text).to(cast_dtype)  # [batch_size, n_ctx, d_model]
+
+        x = x + m.positional_embedding.to(cast_dtype)
+        x = x.permute(1, 0, 2)  # NLD -> LND
+        x = m.transformer(x, attn_mask=m.attn_mask)
+        x = x.permute(1, 0, 2)  # LND -> NLD
+        x = m.ln_final(x)  # [batch_size, n_ctx, transformer.width]
+
+        xx = x @ m.text_projection
+
+        x = x[torch.arange(x.shape[0]), text.argmax(dim=-1)] @ m.text_projection
+
+        return (
+            F.normalize(x, dim=-1) if normalize else x,
+            F.normalize(xx, dim=-1) if normalize else xx,
+        )

approach/ovod/APE/ape/modeling/text/eva01_clip/README.md

@@ -0,0 +1,79 @@
+# Contrastive Language-Image Pre-Training with EVA (EVA-CLIP)
+
+**Table of Contents**
+
+- [Contrastive Language-Image Pre-Training with EVA (EVA-CLIP)](#contrastive-language-image-pre-training-with-eva-eva-clip)
+  - [Model Card](#model-card)
+  - [Usage](#usage)
+  - [Acknowledgement](#acknowledgement)
+  
+
+## Model Card
+
+<div align="center">
+
+| model name | #param. | precision | data  |  batch size | IN-1K zero-shot top-1 | weight |
+|:-----------:|:------:|:------:|:------:|:------:|:------:|:------:|
+| `eva_clip_psz14` | 1.3B | `fp16` | [LAION-400M](https://laion.ai/laion-400-open-dataset/) | 41K | 78.5 | [🤗 HF link](https://huggingface.co/BAAI/EVA/blob/main/eva_clip_psz14.pt) (`2GB`) |
+
+</div>
+
+> The ImageNet-1K zero-shot classification performance is higher than our paper (`78.5` *v.s.* `78.2`) because of longer training.
+
+We choose to train a 1.3B CLIP model, not because it is easy, but because it is hard. Please refer to [this note](https://docs.google.com/document/d/1FXosAZ3wMrzThgnWR6KSkXIz4IMItq3umDGos38pJps/edit) for a glance of the challenges in training very large CLIP.
+
+To our knowledge, EVA-CLIP is **the largest performant open-sourced CLIP model** evaluated via zero-shot classification performance.
+We will updates the results in our paper soon.
+For more details of EVA-CLIP, please refer to Section 2.3.5 of [our paper](https://arxiv.org/pdf/2211.07636.pdf).
+
+We hope open-sourcing EVA-CLIP can facilitate future research in multi-modal learning, representation leaning, AIGC, *etc*.
+
+
+## Usage
+
+The usege of EVA-CLIP is similar to [OpenAI CLIP](https://github.com/openai/CLIP) and [Open CLIP](https://github.com/mlfoundations/open_clip).
+Here we provide a showcase in zero-shot image classification.
+
+First, [install PyTorch 1.7.1](https://pytorch.org/get-started/locally/) (or later) and torchvision, as well as small additional dependencies, and then install this repo as a Python package. On a CUDA GPU machine, the following will do the trick:
+
+```bash
+$ conda install --yes -c pytorch pytorch=1.7.1 torchvision cudatoolkit=11.0
+$ pip install ftfy regex tqdm
+```
+
+The training code of our 1.3B EVA-CLIP will be available at [FlagAI](https://github.com/FlagAI-Open/FlagAI). Please stay tuned.
+
+
+An example:
+```python
+import torch
+from eva_clip import build_eva_model_and_transforms
+from clip import tokenize
+from PIL import Image
+
+eva_clip_path = "/path/to/eva_clip_psz14.pt" # https://huggingface.co/BAAI/EVA/blob/main/eva_clip_psz14.pt
+model_name = "EVA_CLIP_g_14"
+image_path = "CLIP.png"
+caption = ["a diagram", "a dog", "a cat"]
+
+device = "cuda" if torch.cuda.is_available() else "cpu"
+model, preprocess = build_eva_model_and_transforms(model_name, pretrained=eva_clip_path)
+model = model.to(device)
+
+image = preprocess(Image.open(image_path)).unsqueeze(0).to(device)
+text = tokenize(caption).to(device)
+
+with torch.no_grad():
+    image_features = model.encode_image(image)
+    text_features = model.encode_text(text)
+    image_features /= image_features.norm(dim=-1, keepdim=True)
+    text_features /= text_features.norm(dim=-1, keepdim=True)
+
+    text_probs = (100.0 * image_features @ text_features.T).softmax(dim=-1)
+
+print("Label probs:", text_probs)  # prints: [1.0000e+00, 2.0857e-10, 4.8534e-12]
+```
+
+
+## Acknowledgement
+EVA-CLIP is bulit with [OpenAI CLIP](https://github.com/openai/CLIP), [Open CLIP](https://github.com/mlfoundations/open_clip) and [CLIP Benchmark](https://github.com/LAION-AI/CLIP_benchmark). Thanks for their awesome work!

approach/ovod/APE/ape/modeling/text/eva01_clip/__init__.py

@@ -0,0 +1,7 @@
+# from .clip import *
+# from .eva_clip import *
+# from .model import *
+# from .simple_tokenizer import *
+# from .vit_model import *
+
+from .eva_clip import build_eva_model_and_transforms

approach/ovod/APE/ape/modeling/text/eva01_clip/clip.py

@@ -0,0 +1,232 @@
+import hashlib
+import os
+import urllib
+import warnings
+from typing import Any, Union, List
+from pkg_resources import packaging
+
+import torch
+from PIL import Image
+from torchvision.transforms import Compose, Resize, CenterCrop, ToTensor, Normalize
+from tqdm import tqdm
+
+from .model import build_model
+from .simple_tokenizer import SimpleTokenizer as _Tokenizer
+
+try:
+    from torchvision.transforms import InterpolationMode
+    BICUBIC = InterpolationMode.BICUBIC
+except ImportError:
+    BICUBIC = Image.BICUBIC
+
+
+if packaging.version.parse(torch.__version__) < packaging.version.parse("1.7.1"):
+    warnings.warn("PyTorch version 1.7.1 or higher is recommended")
+
+
+__all__ = ["available_models", "load", "tokenize"]
+_tokenizer = _Tokenizer()
+
+_MODELS = {
+    "RN50": "https://openaipublic.azureedge.net/clip/models/afeb0e10f9e5a86da6080e35cf09123aca3b358a0c3e3b6c78a7b63bc04b6762/RN50.pt",
+    "RN101": "https://openaipublic.azureedge.net/clip/models/8fa8567bab74a42d41c5915025a8e4538c3bdbe8804a470a72f30b0d94fab599/RN101.pt",
+    "RN50x4": "https://openaipublic.azureedge.net/clip/models/7e526bd135e493cef0776de27d5f42653e6b4c8bf9e0f653bb11773263205fdd/RN50x4.pt",
+    "RN50x16": "https://openaipublic.azureedge.net/clip/models/52378b407f34354e150460fe41077663dd5b39c54cd0bfd2b27167a4a06ec9aa/RN50x16.pt",
+    "RN50x64": "https://openaipublic.azureedge.net/clip/models/be1cfb55d75a9666199fb2206c106743da0f6468c9d327f3e0d0a543a9919d9c/RN50x64.pt",
+    "ViT-B/32": "https://openaipublic.azureedge.net/clip/models/40d365715913c9da98579312b702a82c18be219cc2a73407c4526f58eba950af/ViT-B-32.pt",
+    "ViT-B/16": "https://openaipublic.azureedge.net/clip/models/5806e77cd80f8b59890b7e101eabd078d9fb84e6937f9e85e4ecb61988df416f/ViT-B-16.pt",
+    "ViT-L/14": "https://openaipublic.azureedge.net/clip/models/b8cca3fd41ae0c99ba7e8951adf17d267cdb84cd88be6f7c2e0eca1737a03836/ViT-L-14.pt",
+    "ViT-L/14@336px": "https://openaipublic.azureedge.net/clip/models/3035c92b350959924f9f00213499208652fc7ea050643e8b385c2dac08641f02/ViT-L-14-336px.pt",
+}
+
+
+def _download(url: str, root: str):
+    os.makedirs(root, exist_ok=True)
+    filename = os.path.basename(url)
+
+    expected_sha256 = url.split("/")[-2]
+    download_target = os.path.join(root, filename)
+
+    if os.path.exists(download_target) and not os.path.isfile(download_target):
+        raise RuntimeError(f"{download_target} exists and is not a regular file")
+
+    if os.path.isfile(download_target):
+        if hashlib.sha256(open(download_target, "rb").read()).hexdigest() == expected_sha256:
+            return download_target
+        else:
+            warnings.warn(f"{download_target} exists, but the SHA256 checksum does not match; re-downloading the file")
+
+    with urllib.request.urlopen(url) as source, open(download_target, "wb") as output:
+        with tqdm(total=int(source.info().get("Content-Length")), ncols=80, unit='iB', unit_scale=True, unit_divisor=1024) as loop:
+            while True:
+                buffer = source.read(8192)
+                if not buffer:
+                    break
+
+                output.write(buffer)
+                loop.update(len(buffer))
+
+    if hashlib.sha256(open(download_target, "rb").read()).hexdigest() != expected_sha256:
+        raise RuntimeError(f"Model has been downloaded but the SHA256 checksum does not not match")
+
+    return download_target
+
+
+def _convert_image_to_rgb(image):
+    return image.convert("RGB")
+
+
+def _transform(n_px):
+    return Compose([
+        Resize(n_px, interpolation=BICUBIC),
+        CenterCrop(n_px),
+        _convert_image_to_rgb,
+        ToTensor(),
+        Normalize((0.48145466, 0.4578275, 0.40821073), (0.26862954, 0.26130258, 0.27577711)),
+    ])
+
+
+def available_models() -> List[str]:
+    """Returns the names of available CLIP models"""
+    return list(_MODELS.keys())
+
+
+def load(name: str, device: Union[str, torch.device] = "cuda" if torch.cuda.is_available() else "cpu", jit: bool = False, download_root: str = None):
+    """Load a CLIP model
+
+    Parameters
+    ----------
+    name : str
+        A model name listed by `clip.available_models()`, or the path to a model checkpoint containing the state_dict
+
+    device : Union[str, torch.device]
+        The device to put the loaded model
+
+    jit : bool
+        Whether to load the optimized JIT model or more hackable non-JIT model (default).
+
+    download_root: str
+        path to download the model files; by default, it uses "~/.cache/clip"
+
+    Returns
+    -------
+    model : torch.nn.Module
+        The CLIP model
+
+    preprocess : Callable[[PIL.Image], torch.Tensor]
+        A torchvision transform that converts a PIL image into a tensor that the returned model can take as its input
+    """
+    if name in _MODELS:
+        model_path = _download(_MODELS[name], download_root or os.path.expanduser("~/.cache/clip"))
+    elif os.path.isfile(name):
+        model_path = name
+    else:
+        raise RuntimeError(f"Model {name} not found; available models = {available_models()}")
+
+    try:
+        # loading JIT archive
+        model = torch.jit.load(model_path, map_location=device if jit else "cpu").eval()
+        state_dict = None
+    except RuntimeError:
+        # loading saved state dict
+        if jit:
+            warnings.warn(f"File {model_path} is not a JIT archive. Loading as a state dict instead")
+            jit = False
+        state_dict = torch.load(model_path, map_location="cpu")
+
+    if not jit:
+        model = build_model(state_dict or model.state_dict()).to(device)
+        if str(device) == "cpu":
+            model.float()
+        return model, _transform(model.visual.input_resolution)
+
+    # patch the device names
+    device_holder = torch.jit.trace(lambda: torch.ones([]).to(torch.device(device)), example_inputs=[])
+    device_node = [n for n in device_holder.graph.findAllNodes("prim::Constant") if "Device" in repr(n)][-1]
+
+    def patch_device(module):
+        try:
+            graphs = [module.graph] if hasattr(module, "graph") else []
+        except RuntimeError:
+            graphs = []
+
+        if hasattr(module, "forward1"):
+            graphs.append(module.forward1.graph)
+
+        for graph in graphs:
+            for node in graph.findAllNodes("prim::Constant"):
+                if "value" in node.attributeNames() and str(node["value"]).startswith("cuda"):
+                    node.copyAttributes(device_node)
+
+    model.apply(patch_device)
+    patch_device(model.encode_image)
+    patch_device(model.encode_text)
+
+    # patch dtype to float32 on CPU
+    if str(device) == "cpu":
+        float_holder = torch.jit.trace(lambda: torch.ones([]).float(), example_inputs=[])
+        float_input = list(float_holder.graph.findNode("aten::to").inputs())[1]
+        float_node = float_input.node()
+
+        def patch_float(module):
+            try:
+                graphs = [module.graph] if hasattr(module, "graph") else []
+            except RuntimeError:
+                graphs = []
+
+            if hasattr(module, "forward1"):
+                graphs.append(module.forward1.graph)
+
+            for graph in graphs:
+                for node in graph.findAllNodes("aten::to"):
+                    inputs = list(node.inputs())
+                    for i in [1, 2]:  # dtype can be the second or third argument to aten::to()
+                        if inputs[i].node()["value"] == 5:
+                            inputs[i].node().copyAttributes(float_node)
+
+        model.apply(patch_float)
+        patch_float(model.encode_image)
+        patch_float(model.encode_text)
+
+        model.float()
+
+    return model, _transform(model.input_resolution.item())
+
+
+def tokenize(texts: Union[str, List[str]], context_length: int = 77, truncate: bool = False) -> torch.LongTensor:
+    """
+    Returns the tokenized representation of given input string(s)
+
+    Parameters
+    ----------
+    texts : Union[str, List[str]]
+        An input string or a list of input strings to tokenize
+
+    context_length : int
+        The context length to use; all CLIP models use 77 as the context length
+
+    truncate: bool
+        Whether to truncate the text in case its encoding is longer than the context length
+
+    Returns
+    -------
+    A two-dimensional tensor containing the resulting tokens, shape = [number of input strings, context_length]
+    """
+    if isinstance(texts, str):
+        texts = [texts]
+
+    sot_token = _tokenizer.encoder["<|startoftext|>"]
+    eot_token = _tokenizer.encoder["<|endoftext|>"]
+    all_tokens = [[sot_token] + _tokenizer.encode(text) + [eot_token] for text in texts]
+    result = torch.zeros(len(all_tokens), context_length, dtype=torch.long)
+
+    for i, tokens in enumerate(all_tokens):
+        if len(tokens) > context_length:
+            if truncate:
+                tokens = tokens[:context_length]
+                tokens[-1] = eot_token
+            else:
+                raise RuntimeError(f"Input {texts[i]} is too long for context length {context_length}")
+        result[i, :len(tokens)] = torch.tensor(tokens)
+
+    return result

approach/ovod/APE/ape/modeling/text/eva01_clip/eva_clip.py

@@ -0,0 +1,173 @@
+import json
+import logging
+import os
+import pathlib
+import re
+from copy import deepcopy
+from pathlib import Path
+from tkinter import E
+from typing import Optional, Tuple, Any, Union, List
+
+import torch
+from torchvision.transforms import Normalize, Compose, InterpolationMode, ToTensor, Resize, CenterCrop
+
+from .eva_model import EVA_CLIP, convert_weights_to_fp16
+
+OPENAI_DATASET_MEAN = (0.48145466, 0.4578275, 0.40821073)
+OPENAI_DATASET_STD = (0.26862954, 0.26130258, 0.27577711)
+
+_MODEL_CONFIG_PATHS = [Path(__file__).parent / f"model_configs/"]
+_MODEL_CONFIGS = {}  # directory (model_name: config) of model architecture configs
+
+
+def _natural_key(string_):
+    return [int(s) if s.isdigit() else s for s in re.split(r'(\d+)', string_.lower())]
+
+
+def _rescan_model_configs():
+    global _MODEL_CONFIGS
+
+    config_ext = ('.json',)
+    config_files = []
+    for config_path in _MODEL_CONFIG_PATHS:
+        if config_path.is_file() and config_path.suffix in config_ext:
+            config_files.append(config_path)
+        elif config_path.is_dir():
+            for ext in config_ext:
+                config_files.extend(config_path.glob(f'*{ext}'))
+
+    for cf in config_files:
+        with open(cf, 'r') as f:
+            model_cfg = json.load(f)
+            if all(a in model_cfg for a in ('embed_dim', 'vision_cfg', 'text_cfg')):
+                _MODEL_CONFIGS[cf.stem] = model_cfg
+
+    _MODEL_CONFIGS = {k: v for k, v in sorted(_MODEL_CONFIGS.items(), key=lambda x: _natural_key(x[0]))}
+
+
+_rescan_model_configs()  # initial populate of model config registry
+
+def list_models():
+    """ enumerate available model architectures based on config files """
+    return list(_MODEL_CONFIGS.keys())
+
+
+def add_model_config(path):
+    """ add model config path or file and update registry """
+    if not isinstance(path, Path):
+        path = Path(path)
+    _MODEL_CONFIG_PATHS.append(path)
+    _rescan_model_configs()
+
+def get_model_config(model_name):
+    if model_name in _MODEL_CONFIGS:
+        return deepcopy(_MODEL_CONFIGS[model_name])
+    else:
+        return None
+
+def load_state_dict(checkpoint_path: str, map_location: str='cpu', model_key='model|module|state_dict'):
+    checkpoint = torch.load(checkpoint_path, map_location=map_location)
+
+    for mk in model_key.split('|'):
+        if isinstance(checkpoint, dict) and mk in checkpoint:
+            state_dict = checkpoint[mk]
+            break
+        else:
+            state_dict = checkpoint
+    if next(iter(state_dict.items()))[0].startswith('module'):
+        state_dict = {k[7:]: v for k, v in state_dict.items()}
+    return state_dict
+
+def load_checkpoint(model, checkpoint_path, model_key="model|module|state_dict", strict=True):
+    state_dict = load_state_dict(checkpoint_path, model_key=model_key)
+    incompatible_keys = model.load_state_dict(state_dict, strict=strict)
+    print(incompatible_keys)
+    return incompatible_keys
+
+def create_model(
+        model_name: str,
+        pretrained: str = '',
+        precision: str = 'fp32',
+        device: torch.device = torch.device('cpu'),
+        force_quick_gelu: bool = False,
+):
+    model_name = model_name.replace('/', '-')  # for callers using old naming with / in ViT names
+
+    if model_name in _MODEL_CONFIGS:
+        logging.info(f'Loading {model_name} model config.')
+        model_cfg = deepcopy(_MODEL_CONFIGS[model_name])
+    else:
+        logging.error(f'Model config for {model_name} not found; available models {list_models()}.')
+        raise RuntimeError(f'Model config for {model_name} not found.')
+
+    if force_quick_gelu:
+        # override for use of QuickGELU on non-OpenAI transformer models
+        model_cfg["quick_gelu"] = True
+
+    model = EVA_CLIP(**model_cfg)
+
+    if pretrained:
+        load_checkpoint(model, pretrained)
+                
+    model.to(device=device)
+    if precision == "fp16":
+        assert device.type != 'cpu'
+        convert_weights_to_fp16(model)
+
+    # set image / mean metadata from pretrained_cfg if available, or use default
+    model.visual.image_mean = OPENAI_DATASET_MEAN
+    model.visual.image_std = OPENAI_DATASET_STD
+
+    return model
+
+def _convert_to_rgb(image):
+    return image.convert('RGB')
+
+def image_transform(
+        image_size: int,
+        mean: Optional[Tuple[float, ...]] = None,
+        std: Optional[Tuple[float, ...]] = None,
+):
+    mean = mean or OPENAI_DATASET_MEAN
+    if not isinstance(mean, (list, tuple)):
+        mean = (mean,) * 3
+
+    std = std or OPENAI_DATASET_STD
+    if not isinstance(std, (list, tuple)):
+        std = (std,) * 3
+    
+    if isinstance(image_size, (list, tuple)) and image_size[0] == image_size[1]:
+        # for square size, pass size as int so that Resize() uses aspect preserving shortest edge
+        image_size = image_size[0]
+
+    normalize = Normalize(mean=mean, std=std)
+    
+    transforms = [
+        Resize(image_size, interpolation=InterpolationMode.BICUBIC),
+        CenterCrop(image_size),
+    ]
+    transforms.extend([
+        _convert_to_rgb,
+        ToTensor(),
+        normalize,
+    ])
+    return Compose(transforms)
+
+def build_eva_model_and_transforms(
+        model_name: str,
+        pretrained: str = '',
+        precision: str = 'fp32',
+        device: torch.device = torch.device('cpu'),
+        force_quick_gelu: bool = False,
+        image_mean: Optional[Tuple[float, ...]] = None,
+        image_std: Optional[Tuple[float, ...]] = None,
+):
+    model = create_model(
+        model_name, pretrained, precision, device,
+        force_quick_gelu=force_quick_gelu)
+
+    image_mean = image_mean or getattr(model.visual, 'image_mean', None)
+    image_std = image_std or getattr(model.visual, 'image_std', None)
+    preprocess_val = image_transform(model.visual.image_size, mean=image_mean, std=image_std)
+
+    return model, preprocess_val

approach/ovod/APE/ape/modeling/text/eva01_clip/eva_model.py

@@ -0,0 +1,368 @@
+""" CLIP Model
+
+Adapted from https://github.com/mlfoundations/open_clip
+
+"""
+import math
+from dataclasses import dataclass
+from typing import Tuple, Union, Callable, Optional
+from functools import partial
+import numpy as np
+from collections import OrderedDict
+
+import torch
+import torch.nn.functional as F
+from torch import nn
+
+from .vit_model import VisionTransformer
+
+try:
+    from apex.normalization import FusedLayerNorm
+except:
+    pass
+
+
+class LayerNorm(nn.LayerNorm):
+    """Subclass torch's LayerNorm to handle fp16."""
+
+    def forward(self, x: torch.Tensor):
+        orig_type = x.dtype
+        x = F.layer_norm(x, self.normalized_shape, self.weight, self.bias, self.eps)
+        return x.to(orig_type)
+
+
+class QuickGELU(nn.Module):
+    # NOTE This is slower than nn.GELU or nn.SiLU and uses more GPU memory
+    def forward(self, x: torch.Tensor):
+        return x * torch.sigmoid(1.702 * x)
+
+
+class Attention(nn.Module):
+    def __init__(
+            self,
+            dim,
+            num_heads=8,
+            qkv_bias=True,
+            scaled_cosine=False,
+            scale_heads=False,
+            logit_scale_max=math.log(1. / 0.01),
+            attn_drop=0.,
+            proj_drop=0.
+    ):
+        super().__init__()
+        self.scaled_cosine = scaled_cosine
+        self.scale_heads = scale_heads
+        assert dim % num_heads == 0, 'dim should be divisible by num_heads'
+        self.num_heads = num_heads
+        self.head_dim = dim // num_heads
+        self.scale = self.head_dim ** -0.5
+        self.logit_scale_max = logit_scale_max
+
+        # keeping in_proj in this form (instead of nn.Linear) to match weight scheme of original
+        self.in_proj_weight = nn.Parameter(torch.randn((dim * 3, dim)) * self.scale)
+        if qkv_bias:
+            self.in_proj_bias = nn.Parameter(torch.zeros(dim * 3))
+        else:
+            self.in_proj_bias = None
+
+        if self.scaled_cosine:
+            self.logit_scale = nn.Parameter(torch.log(10 * torch.ones((num_heads, 1, 1))))
+        else:
+            self.logit_scale = None
+        self.attn_drop = nn.Dropout(attn_drop)
+        if self.scale_heads:
+            self.head_scale = nn.Parameter(torch.ones((num_heads, 1, 1)))
+        else:
+            self.head_scale = None
+        self.out_proj = nn.Linear(dim, dim)
+        self.out_drop = nn.Dropout(proj_drop)
+
+    def forward(self, x, attn_mask: Optional[torch.Tensor] = None):
+        L, N, C = x.shape
+        q, k, v = F.linear(x, self.in_proj_weight, self.in_proj_bias).chunk(3, dim=-1)
+        q = q.contiguous().view(L, N * self.num_heads, -1).transpose(0, 1)
+        k = k.contiguous().view(L, N * self.num_heads, -1).transpose(0, 1)
+        v = v.contiguous().view(L, N * self.num_heads, -1).transpose(0, 1)
+
+        if self.logit_scale is not None:
+            attn = torch.bmm(F.normalize(q, dim=-1), F.normalize(k, dim=-1).transpose(-1, -2))
+            logit_scale = torch.clamp(self.logit_scale, max=self.logit_scale_max).exp()
+            attn = attn.view(N, self.num_heads, L, L) * logit_scale
+            attn = attn.view(-1, L, L)
+        else:
+            q = q * self.scale
+            attn = torch.bmm(q, k.transpose(-1, -2))
+
+        if attn_mask is not None:
+            if attn_mask.dtype == torch.bool:
+                new_attn_mask = torch.zeros_like(attn_mask, dtype=q.dtype)
+                new_attn_mask.masked_fill_(attn_mask, float("-inf"))
+                attn_mask = new_attn_mask
+            attn += attn_mask
+
+        attn = attn.softmax(dim=-1)
+        attn = self.attn_drop(attn)
+
+        x = torch.bmm(attn, v)
+        if self.head_scale is not None:
+            x = x.view(N, self.num_heads, L, C) * self.head_scale
+            x = x.view(-1, L, C)
+        x = x.transpose(0, 1).reshape(L, N, C)
+        x = self.out_proj(x)
+        x = self.out_drop(x)
+        return x
+
+
+class ResidualAttentionBlock(nn.Module):
+    def __init__(
+            self,
+            d_model: int,
+            n_head: int,
+            mlp_ratio: float = 4.0,
+            act_layer: Callable = nn.GELU,
+            scale_cosine_attn: bool = False,
+            scale_heads: bool = False,
+            scale_attn: bool = False,
+            scale_fc: bool = False,
+    ):
+        super().__init__()
+
+        self.ln_1 = LayerNorm(d_model)
+        # FIXME torchscript issues need to be resolved for custom attention
+        # if scale_cosine_attn or scale_heads:
+        #     self.attn = Attention(
+        #        d_model, n_head,
+        #        scaled_cosine=scale_cosine_attn,
+        #        scale_heads=scale_heads,
+        #     )
+        self.attn = nn.MultiheadAttention(d_model, n_head)
+        self.ln_attn = LayerNorm(d_model) if scale_attn else nn.Identity()
+
+        self.ln_2 = LayerNorm(d_model)
+        mlp_width = int(d_model * mlp_ratio)
+        self.mlp = nn.Sequential(OrderedDict([
+            ("c_fc", nn.Linear(d_model, mlp_width)),
+            ('ln', LayerNorm(mlp_width) if scale_fc else nn.Identity()),
+            ("gelu", act_layer()),
+            ("c_proj", nn.Linear(mlp_width, d_model))
+        ]))
+
+    def attention(self, x: torch.Tensor, attn_mask: Optional[torch.Tensor] = None):
+        return self.attn(x, x, x, need_weights=False, attn_mask=attn_mask)[0]
+        # FIXME torchscript issues need resolving for custom attention option to work
+        # if self.use_torch_attn:
+        #     return self.attn(x, x, x, need_weights=False, attn_mask=attn_mask)[0]
+        # else:
+        #     return self.attn(x, attn_mask=attn_mask)
+
+    def cross_attention(self, x: torch.Tensor, context: torch.Tensor, attn_mask: Optional[torch.Tensor] = None):
+        return self.attn(x, context, context, need_weights=False, attn_mask=attn_mask)[0]
+
+
+    def forward(self, x: torch.Tensor, attn_mask: Optional[torch.Tensor] = None):
+        x = x + self.ln_attn(self.attention(self.ln_1(x), attn_mask=attn_mask))
+        x = x + self.mlp(self.ln_2(x))
+        return x
+
+class Transformer(nn.Module):
+    def __init__(self, width: int, layers: int, heads: int,  mlp_ratio: float = 4.0, act_layer: Callable = nn.GELU):
+        super().__init__()
+        self.width = width
+        self.layers = layers
+
+        self.resblocks = nn.ModuleList([
+            ResidualAttentionBlock(width, heads, mlp_ratio, act_layer=act_layer)
+            for _ in range(layers)
+        ])
+
+    def forward(self, x: torch.Tensor, attn_mask: Optional[torch.Tensor] = None):
+        for r in self.resblocks:
+            x = r(x, attn_mask=attn_mask)
+        return x
+
+class TextTransformer(nn.Module):
+    def __init__(
+            self,
+            vocab_size: int,
+            width: int,
+            layers: int,
+            heads: int,
+            context_length: int,
+            embed_dim: int,
+            act_layer: Callable = nn.GELU,
+    ):
+        super().__init__()
+        self.transformer = Transformer(
+            width=width,
+            layers=layers,
+            heads=heads,
+            act_layer=act_layer,
+        )
+        self.context_length = context_length
+        self.vocab_size = vocab_size
+        self.token_embedding = nn.Embedding(vocab_size, width)
+        self.positional_embedding = nn.Parameter(torch.empty(context_length, width))
+        self.ln_final = LayerNorm(width)
+
+        self.text_projection = nn.Parameter(torch.empty(width, embed_dim))
+        self.logit_scale = nn.Parameter(torch.ones([]) * np.log(1 / 0.07))
+        self.register_buffer('attn_mask', self.build_attention_mask(), persistent=False)
+
+        self.init_parameters()
+
+    def init_parameters(self):
+        nn.init.normal_(self.token_embedding.weight, std=0.02)
+        nn.init.normal_(self.positional_embedding, std=0.01)
+        nn.init.constant_(self.logit_scale, np.log(1 / 0.07))
+
+        proj_std = (self.transformer.width ** -0.5) * ((2 * self.transformer.layers) ** -0.5)
+        attn_std = self.transformer.width ** -0.5
+        fc_std = (2 * self.transformer.width) ** -0.5
+        for block in self.transformer.resblocks:
+            nn.init.normal_(block.attn.in_proj_weight, std=attn_std)
+            nn.init.normal_(block.attn.out_proj.weight, std=proj_std)
+            nn.init.normal_(block.mlp.c_fc.weight, std=fc_std)
+            nn.init.normal_(block.mlp.c_proj.weight, std=proj_std)
+
+        if self.text_projection is not None:
+            nn.init.normal_(self.text_projection, std=self.transformer.width ** -0.5)
+
+    def build_attention_mask(self):
+        # lazily create causal attention mask, with full attention between the vision tokens
+        # pytorch uses additive attention mask; fill with -inf
+        mask = torch.empty(self.context_length, self.context_length)
+        mask.fill_(float("-inf"))
+        mask.triu_(1)  # zero out the lower diagonal
+        return mask
+
+    def forward_features(self, text: torch.Tensor):
+        x = self.token_embedding(text)  # [batch_size, n_ctx, d_model]
+
+        x = x + self.positional_embedding
+        x = x.permute(1, 0, 2)  # NLD -> LND
+        x = self.transformer(x, attn_mask=self.attn_mask)
+        x = x.permute(1, 0, 2)  # LND -> NLD
+        x = self.ln_final(x)
+
+        # x.shape = [batch_size, n_ctx, transformer.width]
+        # take features from the eot embedding (eot_token is the highest number in each sequence)
+        x = x[torch.arange(x.shape[0]), text.argmax(dim=-1)]
+        return x
+
+    def forward(self, x: torch.Tensor):
+        x = self.forward_features(x)
+        if self.text_projection is not None:
+            x = x @ self.text_projection
+        return x
+
+@dataclass
+class CLIPVisionCfg:
+    layers: Union[Tuple[int, int, int, int], int] = 12
+    width: int = 768
+    head_width: int = 64
+    mlp_ratio: float = 4.0
+    patch_size: int = 16
+    image_size: Union[Tuple[int, int], int] = 224
+    layer_scale_init_value: float = None
+    drop_path_rate: float = 0.
+    fusedLN: bool = False
+    xattn: bool = False
+
+
+@dataclass
+class CLIPTextCfg:
+    context_length: int = 77
+    vocab_size: int = 49408
+    width: int = 512
+    heads: int = 8
+    layers: int = 12
+
+
+
+class EVA_CLIP(nn.Module):
+    def __init__(
+            self,
+            embed_dim: int,
+            vision_cfg: CLIPVisionCfg,
+            text_cfg: CLIPTextCfg,
+            quick_gelu: bool = False,
+    ):
+        super().__init__()
+        if isinstance(vision_cfg, dict):
+            vision_cfg = CLIPVisionCfg(**vision_cfg)
+        if isinstance(text_cfg, dict):
+            text_cfg = CLIPTextCfg(**text_cfg)
+
+        # OpenAI models are pretrained w/ QuickGELU but native nn.GELU is both faster and more
+        # memory efficient in recent PyTorch releases (>= 1.10).
+        act_layer = QuickGELU if quick_gelu else nn.GELU
+
+        vision_heads = vision_cfg.width // vision_cfg.head_width
+        self.visual = VisionTransformer(
+            img_size=vision_cfg.image_size,
+            patch_size=vision_cfg.patch_size,
+            num_classes=embed_dim,
+            use_mean_pooling=False,
+            init_values=vision_cfg.layer_scale_init_value,
+            embed_dim=vision_cfg.width,
+            depth=vision_cfg.layers,
+            num_heads=vision_heads,
+            mlp_ratio=vision_cfg.mlp_ratio,
+            qkv_bias=True,
+            drop_path_rate=vision_cfg.drop_path_rate,
+            norm_layer= partial(FusedLayerNorm, eps=1e-6) if vision_cfg.fusedLN else partial(nn.LayerNorm, eps=1e-6),
+            xattn=vision_cfg.xattn
+        )
+
+        self.text = TextTransformer(
+            vocab_size=text_cfg.vocab_size,
+            width=text_cfg.width,
+            layers=text_cfg.layers,
+            heads=text_cfg.heads,
+            context_length=text_cfg.context_length,
+            embed_dim=embed_dim,
+            act_layer=act_layer
+        )
+
+    def encode_image(self, image):
+        return self.visual(image)
+
+    def encode_text(self, text):
+        return self.text(text)
+
+    def forward(self, image, text):
+        if image is None:
+            return self.encode_text(text)
+        elif text is None:
+            return self.encode_image(image)
+        image_features = self.encode_image(image)
+        image_features = F.normalize(image_features, dim=-1)
+
+        text_features = self.encode_text(text)
+        text_features = F.normalize(text_features, dim=-1)
+
+        return image_features, text_features, self.text.logit_scale.exp()
+
+
+def convert_weights_to_fp16(model: nn.Module):
+    """Convert applicable model parameters to fp16"""
+
+    def _convert_weights_to_fp16(l):
+        if isinstance(l, (nn.Conv1d, nn.Conv2d, nn.Linear)):
+            l.weight.data = l.weight.data.half()
+            if l.bias is not None:
+                l.bias.data = l.bias.data.half()
+
+        if isinstance(l, (nn.MultiheadAttention, Attention)):
+            for attr in [*[f"{s}_proj_weight" for s in ["in", "q", "k", "v"]], "in_proj_bias", "bias_k", "bias_v"]:
+                tensor = getattr(l, attr)
+                if tensor is not None:
+                    tensor.data = tensor.data.half()
+
+        for name in ["text_projection", "proj"]:
+            if hasattr(l, name):
+                attr = getattr(l, name)
+                if attr is not None:
+                    attr.data = attr.data.half()
+
+    model.apply(_convert_weights_to_fp16)

approach/ovod/APE/ape/modeling/text/eva01_clip/model.py

@@ -0,0 +1,471 @@
+from collections import OrderedDict
+from typing import Tuple, Union
+
+import numpy as np
+import torch
+import math
+import torch.nn.functional as F
+from torch import nn
+
+class Bottleneck(nn.Module):
+    expansion = 4
+
+    def __init__(self, inplanes, planes, stride=1):
+        super().__init__()
+
+        # all conv layers have stride 1. an avgpool is performed after the second convolution when stride > 1
+        self.conv1 = nn.Conv2d(inplanes, planes, 1, bias=False)
+        self.bn1 = nn.BatchNorm2d(planes)
+
+        self.conv2 = nn.Conv2d(planes, planes, 3, padding=1, bias=False)
+        self.bn2 = nn.BatchNorm2d(planes)
+
+        self.avgpool = nn.AvgPool2d(stride) if stride > 1 else nn.Identity()
+
+        self.conv3 = nn.Conv2d(planes, planes * self.expansion, 1, bias=False)
+        self.bn3 = nn.BatchNorm2d(planes * self.expansion)
+
+        self.relu = nn.ReLU(inplace=True)
+        self.downsample = None
+        self.stride = stride
+
+        if stride > 1 or inplanes != planes * Bottleneck.expansion:
+            # downsampling layer is prepended with an avgpool, and the subsequent convolution has stride 1
+            self.downsample = nn.Sequential(OrderedDict([
+                ("-1", nn.AvgPool2d(stride)),
+                ("0", nn.Conv2d(inplanes, planes * self.expansion, 1, stride=1, bias=False)),
+                ("1", nn.BatchNorm2d(planes * self.expansion))
+            ]))
+
+    def forward(self, x: torch.Tensor):
+        identity = x
+
+        out = self.relu(self.bn1(self.conv1(x)))
+        out = self.relu(self.bn2(self.conv2(out)))
+        out = self.avgpool(out)
+        out = self.bn3(self.conv3(out))
+
+        if self.downsample is not None:
+            identity = self.downsample(x)
+
+        out += identity
+        out = self.relu(out)
+        return out
+
+
+class AttentionPool2d(nn.Module):
+    def __init__(self, spacial_dim: int, embed_dim: int, num_heads: int, output_dim: int = None):
+        super().__init__()
+        self.positional_embedding = nn.Parameter(torch.randn(spacial_dim ** 2 + 1, embed_dim) / embed_dim ** 0.5)
+        self.k_proj = nn.Linear(embed_dim, embed_dim)
+        self.q_proj = nn.Linear(embed_dim, embed_dim)
+        self.v_proj = nn.Linear(embed_dim, embed_dim)
+        self.c_proj = nn.Linear(embed_dim, output_dim or embed_dim)
+        self.num_heads = num_heads
+
+    def forward(self, x, return_all_tokens=False):
+        x = x.reshape(x.shape[0], x.shape[1], x.shape[2] * x.shape[3]).permute(2, 0, 1)  # NCHW -> (HW)NC
+        x = torch.cat([x.mean(dim=0, keepdim=True), x], dim=0)  # (HW+1)NC
+        x = x + self.positional_embedding[:, None, :].to(x.dtype)  # (HW+1)NC
+        x, _ = F.multi_head_attention_forward(
+            query=x, key=x, value=x,
+            embed_dim_to_check=x.shape[-1],
+            num_heads=self.num_heads,
+            q_proj_weight=self.q_proj.weight,
+            k_proj_weight=self.k_proj.weight,
+            v_proj_weight=self.v_proj.weight,
+            in_proj_weight=None,
+            in_proj_bias=torch.cat([self.q_proj.bias, self.k_proj.bias, self.v_proj.bias]),
+            bias_k=None,
+            bias_v=None,
+            add_zero_attn=False,
+            dropout_p=0,
+            out_proj_weight=self.c_proj.weight,
+            out_proj_bias=self.c_proj.bias,
+            use_separate_proj_weight=True,
+            training=self.training,
+            need_weights=False
+        )
+        if return_all_tokens:
+            return x
+        else:
+            return x[0]
+
+
+class ModifiedResNet(nn.Module):
+    """
+    A ResNet class that is similar to torchvision's but contains the following changes:
+    - There are now 3 "stem" convolutions as opposed to 1, with an average pool instead of a max pool.
+    - Performs anti-aliasing strided convolutions, where an avgpool is prepended to convolutions with stride > 1
+    - The final pooling layer is a QKV attention instead of an average pool
+    """
+
+    def __init__(self, layers, output_dim, heads, input_resolution=224, width=64):
+        super().__init__()
+        self.output_dim = output_dim
+        self.input_resolution = input_resolution
+
+        # the 3-layer stem
+        self.conv1 = nn.Conv2d(3, width // 2, kernel_size=3, stride=2, padding=1, bias=False)
+        self.bn1 = nn.BatchNorm2d(width // 2)
+        self.conv2 = nn.Conv2d(width // 2, width // 2, kernel_size=3, padding=1, bias=False)
+        self.bn2 = nn.BatchNorm2d(width // 2)
+        self.conv3 = nn.Conv2d(width // 2, width, kernel_size=3, padding=1, bias=False)
+        self.bn3 = nn.BatchNorm2d(width)
+        self.avgpool = nn.AvgPool2d(2)
+        self.relu = nn.ReLU(inplace=True)
+
+        # residual layers
+        self._inplanes = width  # this is a *mutable* variable used during construction
+        self.layer1 = self._make_layer(width, layers[0])
+        self.layer2 = self._make_layer(width * 2, layers[1], stride=2)
+        self.layer3 = self._make_layer(width * 4, layers[2], stride=2)
+        self.layer4 = self._make_layer(width * 8, layers[3], stride=2)
+
+        embed_dim = width * 32  # the ResNet feature dimension
+        self.attnpool = AttentionPool2d(input_resolution // 32, embed_dim, heads, output_dim)
+
+    def _make_layer(self, planes, blocks, stride=1):
+        layers = [Bottleneck(self._inplanes, planes, stride)]
+
+        self._inplanes = planes * Bottleneck.expansion
+        for _ in range(1, blocks):
+            layers.append(Bottleneck(self._inplanes, planes))
+
+        return nn.Sequential(*layers)
+
+    def forward(self, x, return_side_out=False, return_all_tokens=False):
+        def stem(x):
+            for conv, bn in [(self.conv1, self.bn1), (self.conv2, self.bn2), (self.conv3, self.bn3)]:
+                x = self.relu(bn(conv(x)))
+            x = self.avgpool(x)
+            return x
+        out = []
+        x = x.type(self.conv1.weight.dtype)
+        x = stem(x)
+        x = self.layer1(x)
+        if return_side_out:
+            out.append(x)
+        x = self.layer2(x)
+        if return_side_out:
+            out.append(x)
+        x = self.layer3(x)
+        if return_side_out:
+            out.append(x)
+        x = self.layer4(x)
+        if return_side_out:
+            out.append(x)
+        x = self.attnpool(x, return_all_tokens)
+        out.append(x)
+        if len(out) == 1:
+            return x
+        else:
+            return out
+
+
+class LayerNorm(nn.LayerNorm):
+    """Subclass torch's LayerNorm to handle fp16."""
+
+    def forward(self, x: torch.Tensor):
+        orig_type = x.dtype
+        ret = super().forward(x.type(torch.float32))
+        return ret.type(orig_type)
+
+
+class QuickGELU(nn.Module):
+    def forward(self, x: torch.Tensor):
+        return x * torch.sigmoid(1.702 * x)
+
+
+class ResidualAttentionBlock(nn.Module):
+    def __init__(self, d_model: int, n_head: int, attn_mask: torch.Tensor = None):
+        super().__init__()
+
+        self.attn = nn.MultiheadAttention(d_model, n_head)
+        self.ln_1 = LayerNorm(d_model)
+        self.mlp = nn.Sequential(OrderedDict([
+            ("c_fc", nn.Linear(d_model, d_model * 4)),
+            ("gelu", QuickGELU()),
+            ("c_proj", nn.Linear(d_model * 4, d_model))
+        ]))
+        self.ln_2 = LayerNorm(d_model)
+        self.attn_mask = attn_mask
+
+    def attention(self, x: torch.Tensor):
+        self.attn_mask = self.attn_mask.to(dtype=x.dtype, device=x.device) if self.attn_mask is not None else None
+        return self.attn(x, x, x, need_weights=False, attn_mask=self.attn_mask)[0]
+
+
+    def forward(self, x: torch.Tensor):
+        x = x + self.attention(self.ln_1(x))
+        x = x + self.mlp(self.ln_2(x))
+        return x
+
+
+class Transformer(nn.Module):
+    def __init__(self, width: int, layers: int, heads: int, attn_mask: torch.Tensor = None):
+        super().__init__()
+        self.width = width
+        self.layers = layers
+        self.resblocks = nn.Sequential(*[ResidualAttentionBlock(width, heads, attn_mask) for _ in range(layers)])
+
+    def forward(self, x: torch.Tensor):
+        return self.resblocks(x)
+
+
+class VisionTransformer(nn.Module):
+    def __init__(self, input_resolution: int, patch_size: int, width: int, layers: int, heads: int, output_dim: int):
+        super().__init__()
+        self.input_resolution = input_resolution
+        self.output_dim = output_dim
+        self.conv1 = nn.Conv2d(in_channels=3, out_channels=width, kernel_size=patch_size, stride=patch_size, bias=False)
+
+        scale = width ** -0.5
+        self.class_embedding = nn.Parameter(scale * torch.randn(width))
+        self.positional_embedding = nn.Parameter(scale * torch.randn((input_resolution // patch_size) ** 2 + 1, width))
+        self.ln_pre = LayerNorm(width)
+        self.patch_shape = (input_resolution // patch_size, ) * 2
+
+        self.patch_size = patch_size
+
+        self.transformer = Transformer(width, layers, heads)
+
+        self.ln_post = LayerNorm(width)
+        self.proj = nn.Parameter(scale * torch.randn(width, output_dim))
+
+    def interpolate_pos_encoding(self, x, w, h):
+        class_pos_embed = self.positional_embedding[0]
+        patch_pos_embed = self.positional_embedding[1:]
+        dim = x.shape[-1]
+        # we add a small number to avoid floating point error in the interpolation
+        # see discussion at https://github.com/facebookresearch/dino/issues/8
+        # w0, h0 = w + 0.1, h + 0.1
+        n, m = self.patch_shape
+        patch_pos_embed = nn.functional.interpolate(
+            patch_pos_embed.reshape(1, n, m, dim).permute(0, 3, 1, 2),
+            scale_factor=(w / n, h / m),
+            mode='bicubic',
+        )
+        assert int(w) == patch_pos_embed.shape[-2] and int(h) == patch_pos_embed.shape[-1]
+        patch_pos_embed = patch_pos_embed.permute(0, 2, 3, 1).view(-1, dim)
+        return torch.cat((class_pos_embed.unsqueeze(0), patch_pos_embed), dim=0)
+
+    def forward(self, x: torch.Tensor):
+        x = self.conv1(x)  # shape = [*, width, grid, grid]
+        bsz, _, w, h = x.size()
+        x = x.reshape(x.shape[0], x.shape[1], -1)  # shape = [*, width, grid ** 2]
+        x = x.permute(0, 2, 1)  # shape = [*, grid ** 2, width]
+        x = torch.cat([self.class_embedding.to(x.dtype) + torch.zeros(x.shape[0], 1, x.shape[-1], dtype=x.dtype, device=x.device), x], dim=1)  # shape = [*, grid ** 2 + 1, width]
+        if w != self.patch_shape[0] or h != self.patch_shape[1]:
+            x = x + self.interpolate_pos_encoding(x, w, h)
+        else:
+            x = x + self.positional_embedding.to(x.dtype)
+        x = self.ln_pre(x)
+
+        x = x.permute(1, 0, 2)  # NLD -> LND
+        x = self.transformer(x)
+        x = x.permute(1, 0, 2)  # LND -> NLD
+
+        x = x[:, 1:, :]             ### w/o cls token
+        x = self.ln_post(x)         ### norm
+        if self.proj is not None:
+            x = x @ self.proj       ### proj to low dim
+        return x
+
+
+
+class CLIP(nn.Module):
+    def __init__(self,
+                 embed_dim: int,
+                 # vision
+                 image_resolution: int,
+                 vision_layers: Union[Tuple[int, int, int, int], int],
+                 vision_width: int,
+                 vision_patch_size: int,
+                 # text
+                 context_length: int,
+                 vocab_size: int,
+                 transformer_width: int,
+                 transformer_heads: int,
+                 transformer_layers: int
+                 ):
+        super().__init__()
+
+        self.context_length = context_length
+
+        if isinstance(vision_layers, (tuple, list)):
+            vision_heads = vision_width * 32 // 64
+            self.visual = ModifiedResNet(
+                layers=vision_layers,
+                output_dim=embed_dim,
+                heads=vision_heads,
+                input_resolution=image_resolution,
+                width=vision_width
+            )
+        else:
+            vision_heads = vision_width // 64
+            self.visual = VisionTransformer(
+                input_resolution=image_resolution,
+                patch_size=vision_patch_size,
+                width=vision_width,
+                layers=vision_layers,
+                heads=vision_heads,
+                output_dim=embed_dim
+            )
+
+        self.transformer = Transformer(
+            width=transformer_width,
+            layers=transformer_layers,
+            heads=transformer_heads,
+            attn_mask=self.build_attention_mask()
+        )
+
+        self.vocab_size = vocab_size
+        self.token_embedding = nn.Embedding(vocab_size, transformer_width)
+        self.positional_embedding = nn.Parameter(torch.empty(self.context_length, transformer_width))
+        self.ln_final = LayerNorm(transformer_width)
+
+        self.text_projection = nn.Parameter(torch.empty(transformer_width, embed_dim))
+        self.logit_scale = nn.Parameter(torch.ones([]) * np.log(1 / 0.07))
+
+        self.initialize_parameters()
+
+    def initialize_parameters(self):
+        nn.init.normal_(self.token_embedding.weight, std=0.02)
+        nn.init.normal_(self.positional_embedding, std=0.01)
+
+        if isinstance(self.visual, ModifiedResNet):
+            if self.visual.attnpool is not None:
+                std = self.visual.attnpool.c_proj.in_features ** -0.5
+                nn.init.normal_(self.visual.attnpool.q_proj.weight, std=std)
+                nn.init.normal_(self.visual.attnpool.k_proj.weight, std=std)
+                nn.init.normal_(self.visual.attnpool.v_proj.weight, std=std)
+                nn.init.normal_(self.visual.attnpool.c_proj.weight, std=std)
+
+            for resnet_block in [self.visual.layer1, self.visual.layer2, self.visual.layer3, self.visual.layer4]:
+                for name, param in resnet_block.named_parameters():
+                    if name.endswith("bn3.weight"):
+                        nn.init.zeros_(param)
+
+        proj_std = (self.transformer.width ** -0.5) * ((2 * self.transformer.layers) ** -0.5)
+        attn_std = self.transformer.width ** -0.5
+        fc_std = (2 * self.transformer.width) ** -0.5
+        for block in self.transformer.resblocks:
+            nn.init.normal_(block.attn.in_proj_weight, std=attn_std)
+            nn.init.normal_(block.attn.out_proj.weight, std=proj_std)
+            nn.init.normal_(block.mlp.c_fc.weight, std=fc_std)
+            nn.init.normal_(block.mlp.c_proj.weight, std=proj_std)
+
+        if self.text_projection is not None:
+            nn.init.normal_(self.text_projection, std=self.transformer.width ** -0.5)
+
+    def build_attention_mask(self):
+        # lazily create causal attention mask, with full attention between the vision tokens
+        # pytorch uses additive attention mask; fill with -inf
+        mask = torch.empty(self.context_length, self.context_length)
+        mask.fill_(float("-inf"))
+        mask.triu_(1)  # zero out the lower diagonal
+        return mask
+
+    @property
+    def dtype(self):
+        return self.visual.conv1.weight.dtype
+
+    def encode_image(self, image):
+        return self.visual(image.type(self.dtype))
+
+    def encode_text(self, text):
+        x = self.token_embedding(text).type(self.dtype)  # [batch_size, n_ctx, d_model]
+
+        x = x + self.positional_embedding.type(self.dtype)
+        x = x.permute(1, 0, 2)  # NLD -> LND
+        x = self.transformer(x)
+        x = x.permute(1, 0, 2)  # LND -> NLD
+        x = self.ln_final(x).type(self.dtype)
+
+        # x.shape = [batch_size, n_ctx, transformer.width]
+        # take features from the eot embedding (eot_token is the highest number in each sequence)
+        x = x[torch.arange(x.shape[0]), text.argmax(dim=-1)] @ self.text_projection
+
+        return x
+
+    def forward(self, image, text):
+        image_features = self.encode_image(image)
+        text_features = self.encode_text(text)
+
+        # normalized features
+        image_features = image_features / image_features.norm(dim=-1, keepdim=True)
+        text_features = text_features / text_features.norm(dim=-1, keepdim=True)
+
+        # cosine similarity as logits
+        logit_scale = self.logit_scale.exp()
+        logits_per_image = logit_scale * image_features @ text_features.t()
+        logits_per_text = logits_per_image.t()
+
+        # shape = [global_batch_size, global_batch_size]
+        return logits_per_image, logits_per_text
+
+
+def convert_weights(model: nn.Module):
+    """Convert applicable model parameters to fp16"""
+
+    def _convert_weights_to_fp16(l):
+        if isinstance(l, (nn.Conv1d, nn.Conv2d, nn.Linear)):
+            l.weight.data = l.weight.data.half()
+            if l.bias is not None:
+                l.bias.data = l.bias.data.half()
+
+        if isinstance(l, nn.MultiheadAttention):
+            for attr in [*[f"{s}_proj_weight" for s in ["in", "q", "k", "v"]], "in_proj_bias", "bias_k", "bias_v"]:
+                tensor = getattr(l, attr)
+                if tensor is not None:
+                    tensor.data = tensor.data.half()
+
+        for name in ["text_projection", "proj"]:
+            if hasattr(l, name):
+                attr = getattr(l, name)
+                if attr is not None:
+                    attr.data = attr.data.half()
+
+    model.apply(_convert_weights_to_fp16)
+
+
+
+def build_model(state_dict: dict):
+    vit = "visual.proj" in state_dict
+
+    if vit:
+        vision_width = state_dict["visual.conv1.weight"].shape[0]
+        vision_layers = len([k for k in state_dict.keys() if k.startswith("visual.") and k.endswith(".attn.in_proj_weight")])
+        vision_patch_size = state_dict["visual.conv1.weight"].shape[-1]
+        grid_size = round((state_dict["visual.positional_embedding"].shape[0] - 1) ** 0.5)
+        image_resolution = vision_patch_size * grid_size
+    else:
+        counts: list = [len(set(k.split(".")[2] for k in state_dict if k.startswith(f"visual.layer{b}"))) for b in [1, 2, 3, 4]]
+        vision_layers = tuple(counts)
+        vision_width = state_dict["visual.layer1.0.conv1.weight"].shape[0]
+        output_width = round((state_dict["visual.attnpool.positional_embedding"].shape[0] - 1) ** 0.5)
+        vision_patch_size = None
+        assert output_width ** 2 + 1 == state_dict["visual.attnpool.positional_embedding"].shape[0]
+        image_resolution = output_width * 32
+
+    embed_dim = state_dict["text_projection"].shape[1]
+    context_length = state_dict["positional_embedding"].shape[0]
+    vocab_size = state_dict["token_embedding.weight"].shape[0]
+    transformer_width = state_dict["ln_final.weight"].shape[0]
+    transformer_heads = transformer_width // 64
+    transformer_layers = len(set(k.split(".")[2] for k in state_dict if k.startswith(f"transformer.resblocks")))
+
+    model = CLIP(
+        embed_dim,
+        image_resolution, vision_layers, vision_width, vision_patch_size,
+        context_length, vocab_size, transformer_width, transformer_heads, transformer_layers
+    )
+
+    for key in ["input_resolution", "context_length", "vocab_size"]:
+        if key in state_dict:
+            del state_dict[key]
+
+    convert_weights(model)
+    model.load_state_dict(state_dict)
+    return model.eval()

approach/ovod/APE/ape/modeling/text/eva01_clip/model_configs/EVA_CLIP_g_14.json

@@ -0,0 +1,19 @@
+{
+    "embed_dim": 1024,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 40,
+        "width": 1408,
+        "head_width": 88,
+        "mlp_ratio": 4.3637,
+        "patch_size": 14,
+        "drop_path_rate": 0.0
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 768,
+        "heads": 12,
+        "layers": 12
+    }
+}