Data for evaluate.py

# Initialize NuScenes object.

# We do not store it in self to let garbage collection take care of it and save memory.

nusc = NuScenes(version=nusc_version, verbose=verbose, dataroot=nusc_dataroot)

nusc가 필요한 부분을 정리

 

1. load.gt()

gt_boxes = load_gt(nusc, self.eval_set, TrackingBox, verbose=verbose)

sample.json['token']

# Read out all sample_tokens in DB.

sample_tokens_all = [s['token'] for s in nusc.sample]

sample.json['scene_token']

scene.json['name']

for sample_token in sample_tokens_all:

scene_token = nusc.get('sample', sample_token)['scene_token']

scene_record = nusc.get('scene', scene_token)

if scene_record['name'] in splits[eval_split]:

sample_tokens.append(sample_token)

 

sample.json['sample_token']

sample_annotation.json

for ann_record in self.sample_annotation:

sample_record = self.get('sample', ann_record['sample_token'])

sample_record['anns'].append(ann_record['token'])

sample = nusc.get('sample', sample_token)

sample_annotation_tokens = sample['anns']

for sample_annotation_token in sample_annotation_tokens:

      sample_annotation = nusc.get('sample_annotation', sample_annotation_token)

 

sample_annotation.json['instance_token']

# Use nuScenes token as tracking id.

tracking_id = sample_annotation['instance_token']

 

instance.json['category_token']

sample_annotation.json['instance_token']

category.json['name']

# Decorate (adds short-cut) sample_annotation table with for category name.

for record in self.sample_annotation:

      inst = self.get('instance', record['instance_token'])

      record['category_name'] = self.get('category', inst['category_token'])['name']

tracking_name = category_to_tracking_name(sample_annotation['category_name'])

 

sample_annotation.json['translation']['size']['rotation']

sample_annotation.json['prev']['next']

sample_annotation.json['num_lidar_pts']['num_radar_pts']

sample.json['timestamp']

sample_boxes.append(

box_cls(

sample_token=sample_token,

translation=sample_annotation['translation'],

size=sample_annotation['size'],

rotation=sample_annotation['rotation'],

velocity=nusc.box_velocity(sample_annotation['token'])[:2],

num_pts=sample_annotation['num_lidar_pts'] + sample_annotation['num_radar_pts'],

tracking_id=tracking_id,

tracking_name=tracking_name,

tracking_score=-1.0 # GT samples do not have a score.

)

)

 

2. add_center_dist()

# Add center distances.

pred_boxes = add_center_dist(nusc, pred_boxes)

gt_boxes = add_center_dist(nusc, gt_boxes)

 

sample.json

sample_data.json['ego_pose_token']

for sample_token in eval_boxes.sample_tokens:

sample_rec = nusc.get('sample', sample_token)

sd_record = nusc.get('sample_data', sample_rec['data']['LIDAR_TOP'])

pose_record = nusc.get('ego_pose', sd_record['ego_pose_token'])

 

ego_pose.json['translation']

ego_translation = (box.translation[0] - pose_record['translation'][0],

box.translation[1] - pose_record['translation'][1],

box.translation[2] - pose_record['translation'][2])

 

3. filter_eval_boxes()

pred_boxes = filter_eval_boxes(nusc, pred_boxes, self.cfg.class_range, verbose=verbose)

gt_boxes = filter_eval_boxes(nusc, gt_boxes, self.cfg.class_range, verbose=verbose)

 

sample_annotation.json

instance.json

category.json

# Perform bike-rack filtering.

sample_anns = nusc.get('sample', sample_token)['anns']

bikerack_recs = [nusc.get('sample_annotation', ann) for ann in sample_anns if

nusc.get('sample_annotation', ann)['category_name'] == 'static_object.bicycle_rack']

 

4. create_tracks()

# Convert boxes to tracks format.

self.tracks_gt = create_tracks(gt_boxes, nusc, self.eval_set, gt=True)

self.tracks_pred = create_tracks(pred_boxes, nusc, self.eval_set, gt=False)

 

sample.json['scene_token']

scene.json

scene_token = nusc.get('sample', sample_token)['scene_token']

scene = nusc.get('scene', scene_token)

 

 

최종

sample.json

sample_data.json

sample_annotation.json

scene.json

instance.json

category.json

ego_pose.json