Skip to content

Commit 9f0cbb5

Browse files
aras62aras62
authored
Interface updated
1 parent 242d879 commit 9f0cbb5

File tree

1 file changed

+79
-67
lines changed

1 file changed

+79
-67
lines changed

‎pie_data.py

Lines changed: 79 additions & 67 deletions
Original file line numberDiff line numberDiff line change
@@ -216,20 +216,22 @@ def _map_text_to_scalar(self, label_type, value):
216216
"""
217217
map_dic = {'occlusion': {'none': 0, 'part': 1, 'full': 2},
218218
'action': {'standing': 0, 'walking': 1},
219-
'nod': {'__undefined__': 0, 'nodding': 1},
220219
'look': {'not-looking': 0, 'looking': 1},
221-
'hand_gesture': {'__undefined__': 0, 'greet': 1, 'yield': 2,
222-
'rightofway': 3, 'other': 4},
223-
'reaction': {'__undefined__': 0, 'clear_path': 1, 'speed_up': 2,
224-
'slow_down': 3},
225-
'cross': {'not-crossing': 0, 'crossing': 1},
220+
'gesture': {'__undefined__': 0, 'hand_ack': 1, 'hand_yield': 2,
221+
'hand_rightofway': 3, 'nod': 4, 'other': 5},
222+
'cross': {'not-crossing': 0, 'crossing': 1, 'crossing-irrelevant': -1},
226223
'age': {'child': 0, 'young': 1, 'adult': 2, 'senior': 3},
227224
'designated': {'ND': 0, 'D': 1},
228225
'gender': {'n/a': 0, 'female': 1, 'male': 2},
229226
'intersection': {'midblock': 0, 'T': 1, 'T-left': 2, 'T-right': 3, 'four-way': 4},
230227
'motion_direction': {'n/a': 0, 'LAT': 1, 'LONG': 2},
231228
'traffic_direction': {'OW': 0, 'TW': 1},
232-
'signalized': {'n/a': 0, 'C': 1, 'S': 2, 'CS': 3}}
229+
'signalized': {'n/a': 0, 'C': 1, 'S': 2, 'CS': 3},
230+
'vehicle': {'car': 0, 'truck': 1, 'bus': 2, 'train': 3, 'bicycle': 4, 'bike': 5},
231+
'sign': {'ped_blue': 0, 'ped_yellow': 1, 'ped_white': 2, 'ped_text': 3, 'stop_sign': 4,
232+
'bus_stop': 5, 'train_stop': 6, 'construction': 7, 'other': 8},
233+
'traffic_light': {'regular': 0, 'transit': 1, 'pedestrian': 2},
234+
'state': {'__undefined__': 0, 'red': 1, 'yellow': 2, 'green': 3}}
233235

234236
return map_dic[label_type][value]
235237

@@ -242,21 +244,23 @@ def _map_scalar_to_text(self, label_type, value):
242244
"""
243245
map_dic = {'occlusion': {0: 'none', 1: 'part', 2: 'full'},
244246
'action': {0: 'standing', 1: 'walking'},
245-
'nod': {0: '__undefined__', 1: 'nodding'},
246247
'look': {0: 'not-looking', 1: 'looking'},
247-
'hand_gesture': {0: '__undefined__', 1: 'greet',
248-
2: 'yield', 3: 'rightofway',
249-
4: 'other'},
250-
'reaction': {0: '__undefined__', 1: 'clear_path',
251-
2: 'speed_up', 3: 'slow_down'},
252-
'cross': {0: 'not-crossing', 1: 'crossing'},
248+
'hand_gesture': {0: '__undefined__', 1: 'hand_ack',
249+
2: 'hand_yield', 3: 'hand_rightofway',
250+
4: 'nod', 5: 'other'},
251+
'cross': {0: 'not-crossing', 1: 'crossing', -1: 'crossing-irrelevant'},
253252
'age': {0: 'child', 1: 'young', 2: 'adult', 3: 'senior'},
254253
'designated': {0: 'ND', 1: 'D'},
255254
'gender': {0: 'n/a', 1: 'female', 2: 'male'},
256255
'intersection': {0: 'midblock', 1: 'T', 2: 'T-left', 3: 'T-right', 4: 'four-way'},
257256
'motion_direction': {0: 'n/a', 1: 'LAT', 2: 'LONG'},
258257
'traffic_direction': {0: 'OW', 1: 'TW'},
259-
'signalized': {0: 'n/a', 1: 'C', 2: 'S', 3: 'CS'}}
258+
'signalized': {0: 'n/a', 1: 'C', 2: 'S', 3: 'CS'},
259+
'vehicle': {0: 'car', 1: 'truck', 2: 'bus', 3: 'train', 4: 'bicycle', 5: 'bike'},
260+
'sign': {0: 'ped_blue', 1: 'ped_yellow', 2: 'ped_white', 3: 'ped_text', 4: 'stop_sign',
261+
5: 'bus_stop', 6: 'train_stop', 7: 'construction', 8: 'other'},
262+
'traffic_light': {0: 'regular', 1: 'transit', 2: 'pedestrian'},
263+
'state': {0: '__undefined__', 1: 'red', 2: 'yellow', 3: 'green'}}
260264

261265
return map_dic[label_type][value]
262266

@@ -272,37 +276,64 @@ def _get_annotations(self, setid, vid):
272276

273277
tree = ET.parse(path_to_file)
274278
ped_annt = 'ped_annotations'
279+
traffic_annt = 'traffic_annotations'
275280

276281
annotations = {}
277282
annotations['num_frames'] = int(tree.find("./meta/task/size").text)
278283
annotations['width'] = int(tree.find("./meta/task/original_size/width").text)
279284
annotations['height'] = int(tree.find("./meta/task/original_size/height").text)
280285
annotations[ped_annt] = {}
286+
annotations[traffic_annt] = {}
281287

282-
ped_tracks = tree.findall("./track")
288+
tracks = tree.findall('./track')
283289

284-
for t in ped_tracks:
290+
for t in tracks:
285291
boxes = t.findall('./box')
286-
ped_id = boxes[0].find('./attribute[@name=\"id\"]').text
287-
annotations[ped_annt][ped_id] = {'frames': [], 'bbox': [], 'occlusion': []}
288-
annotations['ped_annotations'][ped_id]['behavior'] = {'hand_gesture': [],
289-
'look': [],
290-
'action': [],
291-
'nod': []}
292-
293-
for b in boxes:
294-
annotations[ped_annt][ped_id]['bbox'].append(
295-
[float(b.get('xtl')), float(b.get('ytl')),
296-
float(b.get('xbr')), float(b.get('ybr'))])
297-
occ = self._map_text_to_scalar('occlusion',
298-
b.find('./attribute[@name=\"occlusion\"]').text)
299-
annotations[ped_annt][ped_id]['occlusion'].append(occ)
300-
annotations[ped_annt][ped_id]['frames'].append(int(b.get('frame')))
301-
for beh in annotations['ped_annotations'][ped_id]['behavior'].keys():
302-
annotations[ped_annt][ped_id]['behavior'][beh].append(
303-
self._map_text_to_scalar(beh,
304-
b.find('./attribute[@name=\"' + beh + '\"]').text))
292+
obj_label = t.get('label')
293+
obj_id = boxes[0].find('./attribute[@name=\"id\"]').text
294+
295+
if obj_label == 'pedestrian':
296+
annotations[ped_annt][obj_id] = {'frames': [], 'bbox': [], 'occlusion': []}
297+
annotations[ped_annt][obj_id]['behavior'] = {'gesture': [], 'look': [], 'action': [], 'cross': []}
298+
for b in boxes:
299+
# Exclude the annotations that are outside of the frame
300+
if int(b.get('outside')) == 1:
301+
continue
302+
annotations[ped_annt][obj_id]['bbox'].append(
303+
[float(b.get('xtl')), float(b.get('ytl')),
304+
float(b.get('xbr')), float(b.get('ybr'))])
305+
occ = self._map_text_to_scalar('occlusion', b.find('./attribute[@name=\"occlusion\"]').text)
306+
annotations[ped_annt][obj_id]['occlusion'].append(occ)
307+
annotations[ped_annt][obj_id]['frames'].append(int(b.get('frame')))
308+
for beh in annotations['ped_annotations'][obj_id]['behavior']:
309+
# Read behavior tags for each frame and add to the database
310+
annotations[ped_annt][obj_id]['behavior'][beh].append(
311+
self._map_text_to_scalar(beh, b.find('./attribute[@name=\"' + beh + '\"]').text))
305312

313+
else:
314+
obj_type = boxes[0].find('./attribute[@name=\"type\"]')
315+
if obj_type is not None:
316+
obj_type = self._map_text_to_scalar(obj_label,
317+
boxes[0].find('./attribute[@name=\"type\"]').text)
318+
319+
annotations[traffic_annt][obj_id] = {'frames': [], 'bbox': [], 'occlusion': [],
320+
'obj_class': obj_label,
321+
'obj_type': obj_type,
322+
'state': []}
323+
324+
for b in boxes:
325+
# Exclude the annotations that are outside of the frame
326+
if int(b.get('outside')) == 1:
327+
continue
328+
annotations[traffic_annt][obj_id]['bbox'].append(
329+
[float(b.get('xtl')), float(b.get('ytl')),
330+
float(b.get('xbr')), float(b.get('ybr'))])
331+
annotations[traffic_annt][obj_id]['occlusion'].append(int(b.get('occluded')))
332+
annotations[traffic_annt][obj_id]['frames'].append(int(b.get('frame')))
333+
annotations[traffic_annt][obj_id]['frames'].append(int(b.get('frame')))
334+
if obj_label == 'traffic_light':
335+
annotations[traffic_annt][obj_id]['frames'].append(self._map_text_to_scalar('state',
336+
b.find('./attribute[@name=\"state\"]').text))
306337
return annotations
307338

308339
def _get_ped_attributes(self, setid, vid):
@@ -361,16 +392,23 @@ def generate_database(self):
361392
'num_frames': int
362393
'width': int
363394
'height': int
395+
'traffic_annotations'(str): {
396+
'obj_id'(str): {
397+
'frames': list(int)
398+
'occlusion': list(int)
399+
'bbox': list([x1, y1, x2, y2]) (float)
400+
'obj_class': str,
401+
'obj_type': str, # only for traffic lights, vehicles, signs
402+
'state': list(int) # only for traffic lights
364403
'ped_annotations'(str): {
365404
'ped_id'(str): {
366405
'frames': list(int)
367406
'occlusion': list(int)
368407
'bbox': list([x1, y1, x2, y2]) (float)
369408
'behavior'(str): {
370409
'action': list(int)
371-
'reaction': list(int)
372-
'nod': list(int)
373-
'hand_gesture': list(int)
410+
'gesture': list(int)
411+
'cross': list(int)
374412
'look': list(int)
375413
'attributes'(str): {
376414
'age': int
@@ -422,7 +460,6 @@ def generate_database(self):
422460
video_ids = [v.split('_annt.xml')[0] for v in sorted(listdir(join(self._annotation_path,
423461
setid))) if v.endswith("annt.xml")]
424462
database[setid] = {}
425-
426463
for vid in video_ids:
427464
print('Getting annotations for %s, %s' % (setid, vid))
428465
database[setid][vid] = self._get_annotations(setid, vid)
@@ -856,34 +893,9 @@ def _get_crossing(self, image_set, annotations, **params):
856893
num_pedestrians += 1
857894

858895
frame_ids = pid_annots[pid]['frames']
859-
# avoid bounding boxes with width 1 or less
860-
correction_offset = 3
861-
862-
event_frame = pid_annots[pid]['attributes']['crossing_point'] - correction_offset
863-
864-
# # if pid_annots[pid]['attributes']['crossing_point'] < pid_annots[pid]['attributes']['critical_point']:
865-
# if pid_annots[pid]['attributes']['crossing'] < 1 and abs(event_frame - frame_ids[-1]) > 20:
866-
#
867-
# # and \
868-
# # pid_annots[pid]['attributes']['crossing'] < 1:
869-
# print(pid)
870-
# # count += 1
871-
# # print(event_frame)
872-
# # print(frame_ids[-1])
873-
# # print(abs(event_frame - frame_ids[-1]))
874-
# elif pid_annots[pid]['attributes']['crossing'] < 1 and \
875-
# abs(event_frame - frame_ids[-1]) > 20 :
876-
# continue
877-
# # print(pid)
878-
try:
879-
end_idx = frame_ids.index(event_frame)
880-
except:
881-
continue
882-
# print(pid)
883-
# print(event_frame)
884-
# print(frame_ids[0])
885-
# print(frame_ids[-1])
896+
event_frame = pid_annots[pid]['attributes']['crossing_point']
886897

898+
end_idx = frame_ids.index(event_frame)
887899
boxes = pid_annots[pid]['bbox'][:end_idx + 1]
888900
frame_ids = frame_ids[: end_idx + 1]
889901
images = [self._get_image_path(sid, vid, f) for f in frame_ids]

0 commit comments

Comments
 (0)