File size: 6,575 Bytes
729c925
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
# Copyright (c) Aishwarya Kamath & Nicolas Carion. Licensed under the Apache License 2.0. All Rights Reserved
# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved
"""
COCO dataset which returns image_id for evaluation.
Mostly copy-paste from https://github.com/pytorch/vision/blob/13b35ff/references/detection/coco_utils.py
"""
from pathlib import Path

import torch
import torch.utils.data
import torchvision
from pycocotools import mask as coco_mask

import datasets.transforms_image as T


class ModulatedDetection(torchvision.datasets.CocoDetection):
    def __init__(self, img_folder, ann_file, transforms, return_masks):
        super(ModulatedDetection, self).__init__(img_folder, ann_file)
        self._transforms = transforms
        self.prepare = ConvertCocoPolysToMask(return_masks)

    def __getitem__(self, idx):
        instance_check = False
        while not instance_check:
            img, target = super(ModulatedDetection, self).__getitem__(idx)
            image_id = self.ids[idx]
            coco_img = self.coco.loadImgs(image_id)[0]
            caption = coco_img["caption"]
            dataset_name = coco_img["dataset_name"] if "dataset_name" in coco_img else None
            target = {"image_id": image_id, "annotations": target, "caption": caption}
            img, target = self.prepare(img, target)
            if self._transforms is not None:
                img, target = self._transforms(img, target)
            target["dataset_name"] = dataset_name
            for extra_key in ["sentence_id", "original_img_id", "original_id", "task_id"]:
                if extra_key in coco_img:
                    target[extra_key] = coco_img[extra_key] # box xyxy -> cxcywh
            # FIXME: handle "valid", since some box may be removed due to random crop
            target["valid"] = torch.tensor([1]) if len(target["area"]) != 0 else torch.tensor([0])

            if torch.any(target['valid'] == 1):  # at leatst one instance
                instance_check = True
            else:
                import random
                idx = random.randint(0, self.__len__() - 1)
        return img.unsqueeze(0), target
        # return img: [1, 3, H, W], the first dimension means T = 1.


def convert_coco_poly_to_mask(segmentations, height, width):
    masks = []
    for polygons in segmentations:
        rles = coco_mask.frPyObjects(polygons, height, width)
        mask = coco_mask.decode(rles)
        if len(mask.shape) < 3:
            mask = mask[..., None]
        mask = torch.as_tensor(mask, dtype=torch.uint8)
        mask = mask.any(dim=2)
        masks.append(mask)
    if masks:
        masks = torch.stack(masks, dim=0)
    else:
        masks = torch.zeros((0, height, width), dtype=torch.uint8)
    return masks


class ConvertCocoPolysToMask(object):
    def __init__(self, return_masks=False):
        self.return_masks = return_masks

    def __call__(self, image, target):
        w, h = image.size

        image_id = target["image_id"]
        image_id = torch.tensor([image_id])

        anno = target["annotations"]
        caption = target["caption"] if "caption" in target else None

        anno = [obj for obj in anno if "iscrowd" not in obj or obj["iscrowd"] == 0]

        boxes = [obj["bbox"] for obj in anno]
        # guard against no boxes via resizing
        boxes = torch.as_tensor(boxes, dtype=torch.float32).reshape(-1, 4)
        boxes[:, 2:] += boxes[:, :2] # xminyminwh -> xyxy
        boxes[:, 0::2].clamp_(min=0, max=w)
        boxes[:, 1::2].clamp_(min=0, max=h)

        classes = [obj["category_id"] for obj in anno]
        classes = torch.tensor(classes, dtype=torch.int64)

        if self.return_masks:
            segmentations = [obj["segmentation"] for obj in anno]
            masks = convert_coco_poly_to_mask(segmentations, h, w)

        # keep the valid boxes
        keep = (boxes[:, 3] > boxes[:, 1]) & (boxes[:, 2] > boxes[:, 0])
        boxes = boxes[keep]
        classes = classes[keep]
        if self.return_masks:
            masks = masks[keep]

        target = {}
        target["boxes"] = boxes
        target["labels"] = classes
        if caption is not None:
            target["caption"] = caption
        if self.return_masks:
            target["masks"] = masks
        target["image_id"] = image_id

        # for conversion to coco api
        area = torch.tensor([obj["area"] for obj in anno])
        iscrowd = torch.tensor([obj["iscrowd"] if "iscrowd" in obj else 0 for obj in anno])
        target["area"] = area[keep]
        target["iscrowd"] = iscrowd[keep]
        target["valid"] = torch.tensor([1])
        target["orig_size"] = torch.as_tensor([int(h), int(w)])
        target["size"] = torch.as_tensor([int(h), int(w)])
        return image, target


def make_coco_transforms(image_set, cautious):

    normalize = T.Compose([T.ToTensor(), T.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])])

    scales = [480, 512, 544, 576, 608, 640, 672, 704, 736, 768]
    final_scales = [296, 328, 360, 392, 416, 448, 480, 512] 

    max_size = 800
    if image_set == "train":
        horizontal = [] if cautious else [T.RandomHorizontalFlip()]
        return T.Compose(
            horizontal
            + [
                T.RandomSelect(
                    T.RandomResize(scales, max_size=max_size),
                    T.Compose(
                        [
                            T.RandomResize([400, 500, 600]),
                            T.RandomSizeCrop(384, 600, respect_boxes=cautious),
                            T.RandomResize(final_scales, max_size=640),
                        ]
                    ),
                ),
                normalize,
            ]
        )

    if image_set == "val":
        return T.Compose(
            [
                T.RandomResize([360], max_size=640),
                normalize,
            ]
        )

    raise ValueError(f"unknown {image_set}")


def build(dataset_file, image_set, args):
    root = Path(args.coco_path)
    assert root.exists(), f"provided COCO path {root} does not exist"
    mode = "instances"
    dataset = dataset_file
    PATHS = {
        "train": (root / "train2014", root / dataset / f"{mode}_{dataset}_train.json"),
        "val": (root / "train2014", root / dataset / f"{mode}_{dataset}_val.json"),
    }

    img_folder, ann_file = PATHS[image_set]
    dataset = ModulatedDetection(
        img_folder,
        ann_file,
        transforms=make_coco_transforms(image_set, False),
        return_masks=args.masks,
    )
    return dataset