# Copyright (c) Meta Platforms, Inc. and affiliates. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. import csv import logging import os from dataclasses import dataclass from typing import Any, Dict, List, Optional, Set from projectaria_tools.core.sensor_data import TimeDomain, TimeQueryOptions # @manual from projectaria_tools.core.stream_id import StreamId # @manual from .AlignedBox2d import AlignedBox2d from .constants import BOX2D_DATA_CSV_COLUMNS from .io_utils import float_or_none, is_float from .loader_poses_utils import check_csv_columns from .pose_utils import lookup_timestamp logger = logging.getLogger(__name__) logging.basicConfig( level=logging.INFO, format="%(asctime)s - %(filename)s - %(lineno)d - %(levelname)s - %(message)s", ) #物体在图像上的 2D 边界框 @dataclass class ObjectBox2d: box2d: Optional[AlignedBox2d] visibility_ratio: float @dataclass class ObjectBox2dCollection: timestamp_ns: int box2ds: Dict[str, ObjectBox2d] @property def object_uid_list(self) -> Set[str]: return set(self.box2ds.keys()) ObjectBox2dTrajectory = Dict[ int, ObjectBox2dCollection ] # trajectory for a single stream ObjectBox2dTrajectoryCollection = Dict[ str, ObjectBox2dTrajectory ] # trajectories for multiple streams @dataclass class ObjectBox2dCollectionWithDt: box2d_collection: ObjectBox2dCollection time_delta_ns: int class ObjectBox2dProvider: def __init__( self, box2d_trajectory_collection: ObjectBox2dTrajectoryCollection ) -> None: #真正存数据的,但没有进行排序 self._box2d_trajectory_collection = box2d_trajectory_collection #对时间戳进行排序,纯时间序列 self._sorted_timestamp_ns_list: Dict[str, List[int]] = {} for stream_id in self._box2d_trajectory_collection.keys(): self._sorted_timestamp_ns_list[stream_id] = sorted( self._box2d_trajectory_collection[stream_id].keys() ) #生成一个set,里面是所有出现过的物体ID self._object_uids_with_box2ds: Set = { x for box2d_trajectory in self._box2d_trajectory_collection.values() for v in box2d_trajectory.values() for x in v.object_uid_list } def get_timestamp_ns_list(self, stream_id: StreamId) -> Optional[List[int]]: return self._sorted_timestamp_ns_list.get(str(stream_id), None) @property def stream_ids(self) -> List[StreamId]: return [StreamId(x) for x in self._box2d_trajectory_collection.keys()] @property def object_uids(self) -> Set[str]: return set(self._object_uids_with_box2ds) def get_data_statistics(self) -> Dict[str, Any]: """ Returns the stats for Object 2D bounding boxes """ stats = {} stats["num_frames"] = { k: len(v) for k, v in self._sorted_timestamp_ns_list.items() } stats["stream_ids"] = [str(x) for x in self.stream_ids] stats["num_objects"] = len(self.object_uids) stats["object_uids"] = [str(x) for x in self.object_uids] return stats def get_bbox_at_timestamp( self, stream_id: StreamId, timestamp_ns: int, time_query_options: TimeQueryOptions, time_domain: TimeDomain, acceptable_time_delta: Optional[int] = None, ) -> Optional[ObjectBox2dCollectionWithDt]: """ Return the list of boxes at the given timestamp """ if time_domain is not TimeDomain.TIME_CODE: raise ValueError("Value other than TimeDomain.TIME_CODE not yet supported.") if stream_id not in self.stream_ids: raise ValueError(f"Box2d trajectory not available for stream {stream_id}.") box2d_collection, time_delta_ns = lookup_timestamp( time_indexed_dict=self._box2d_trajectory_collection[str(stream_id)], sorted_timestamp_list=self.get_timestamp_ns_list(stream_id=stream_id), query_timestamp=timestamp_ns, time_query_options=time_query_options, ) if ( box2d_collection is None or time_delta_ns is None or ( acceptable_time_delta is not None and abs(time_delta_ns) > acceptable_time_delta ) ): return None else: return ObjectBox2dCollectionWithDt( box2d_collection=box2d_collection, time_delta_ns=time_delta_ns ) # box2d_trajectory_collection[stream_id][timestamp_ns].box2ds[object_uid] = ( # object_box2d # ) def parse_box2ds_from_csv_reader(csv_reader) -> ObjectBox2dTrajectoryCollection: box2d_trajectory_collection: ObjectBox2dTrajectoryCollection = {} # Read the header row header = next(csv_reader) # Ensure we have the desired columns check_csv_columns(header, BOX2D_DATA_CSV_COLUMNS) # Read the rest of the rows in the CSV file for row in csv_reader: stream_id = str(StreamId(row[header.index("stream_id")])) timestamp_ns = int(row[header.index("timestamp[ns]")]) object_uid = str(row[header.index("object_uid")]) visibility_ratio = float_or_none(row[header.index("visibility_ratio[%]")]) #读取和存储已有的框 if is_float(row[header.index("x_min[pixel]")]): x_min_px = float(row[header.index("x_min[pixel]")]) x_max_px = float(row[header.index("x_max[pixel]")]) y_min_px = float(row[header.index("y_min[pixel]")]) y_max_px = float(row[header.index("y_max[pixel]")]) box2d = AlignedBox2d( left=x_min_px, top=y_min_px, right=x_max_px, bottom=y_max_px ) else: box2d = None object_box2d = ObjectBox2d( box2d=box2d, visibility_ratio=visibility_ratio, ) if stream_id not in box2d_trajectory_collection: box2d_trajectory_collection[stream_id] = {} if timestamp_ns not in box2d_trajectory_collection[stream_id]: box2d_trajectory_collection[stream_id][timestamp_ns] = ( ObjectBox2dCollection(timestamp_ns=timestamp_ns, box2ds={}) ) #由于ObjectBox2dCollection用的所对象,所以使用"." box2d_trajectory_collection[stream_id][timestamp_ns].box2ds[object_uid] = ( object_box2d ) return box2d_trajectory_collection def load_box2d_trajectory_from_csv(filename: str) -> Optional[ObjectBox2dProvider]: """Load Objects 2D bounding box meta data from a CSV file. Keyword arguments: filename -- the csv file i.e. sequence_folder + "/box2d_objects.csv" """ if not os.path.exists(filename): logger.warn(f"filename: {filename} does not exist.") return None # Open the CSV file for reading with open(filename, "r") as f: csv_reader = csv.reader(f) box2d_trajectory_collection = parse_box2ds_from_csv_reader( csv_reader=csv_reader ) return ObjectBox2dProvider( box2d_trajectory_collection=box2d_trajectory_collection )