SISTMrL
commented
1 year ago hello, could you please provide the generate_gt_info.py scripts?
Closed wusongyuan closed 1 year ago
SISTMrL
commented
1 year ago hello, could you please provide the generate_gt_info.py scripts?
superkoma
commented
1 year ago 
wusongyuan
commented
1 year ago
- We forgot to delete the instruction of 'generate_gt_info.py' in the tracking part.
- We support to use the gt_info (waymo_infos_train.pkl, waymo_infos_val.pkl) generated in detection module for tracking module currently.
- Please feel free to first generate the gt infos based on instructions of data preprocessing.
Thanks for your reply. When I try to generate infos for tracking, there is an error "KeyError: 'boxes_lidar'". I suppose the reason is that all the annotations in pkl that generate in detection were stored in [''annos']["gt_boxes_lidar"]. However, the script in tracking was trying to parse the annotations in the key named "boxes_lidar".
python -u run_track.py --cfg_file cfgs/tk_model_cfgs/waymo_detzero_track.yaml --data_path ../../data/waymo/waymo_infos_train.pkl --workers 4 --split train
2023-09-20 09:36:02,748 INFO ---------------------------- DetZero Tracking Module ----------------------------
2023-09-20 09:36:02,748 INFO LOCAL_RANK : 0
2023-09-20 09:36:02,748 INFO DATA_CONFIG
2023-09-20 09:36:02,748 INFO DATASET : WaymoTrackDataset
2023-09-20 09:36:02,748 INFO DATA_PATH : ../../data/waymo
2023-09-20 09:36:02,748 INFO PROCESSED_DATA_TAG : waymo_processed_data
2023-09-20 09:36:02,748 INFO CLASS_NAME : ['Vehicle', 'Pedestrian', 'Cyclist']
2023-09-20 09:36:02,748 INFO DATA_PROCESSOR : [{'NAME': 'heading_process'}, {'NAME': 'overlap_box_filter', 'METHOD': 'max_score', 'CLASS_THRESHOLD': {'Vehicle': 0.3, 'Pedestrian': 0.2, 'Cyclist': 0.2}}, {'NAME': 'transform_to_global'}]
2023-09-20 09:36:02,748 INFO _BASECONFIG : cfgs/tk_dataset_cfgs/waymo_dataset.yaml
2023-09-20 09:36:02,748 INFO MODEL
2023-09-20 09:36:02,748 INFO NAME : DetZeroTracker
2023-09-20 09:36:02,748 INFO TRACKING
2023-09-20 09:36:02,748 INFO NAME : TrackManager
2023-09-20 09:36:02,748 INFO FILTER
2023-09-20 09:36:02,748 INFO NAME : KalmanFilter
2023-09-20 09:36:02,748 INFO X_DIM : 5
2023-09-20 09:36:02,748 INFO Z_DIM : 3
2023-09-20 09:36:02,748 INFO Q : [5, 15]
2023-09-20 09:36:02,748 INFO P : [50, 1000]
2023-09-20 09:36:02,748 INFO R : 0.1
2023-09-20 09:36:02,748 INFO DELTA_T : 0.1
2023-09-20 09:36:02,748 INFO TRACK_AGE
2023-09-20 09:36:02,748 INFO BIRTH_AGE : 1
2023-09-20 09:36:02,748 INFO DEATH_AGE : -1
2023-09-20 09:36:02,748 INFO DATA_ASSOCIATION
2023-09-20 09:36:02,748 INFO CLASS_NAME : ['Vehicle', 'Pedestrian', 'Cyclist']
2023-09-20 09:36:02,748 INFO DISTINGUISH_CLASS : True
2023-09-20 09:36:02,748 INFO DISTANCE_METHOD : IoUBEV
2023-09-20 09:36:02,748 INFO ASSIGNMENT_METHOD : GNN
2023-09-20 09:36:02,748 INFO STAGE
2023-09-20 09:36:02,748 INFO NAME : two_stage
2023-09-20 09:36:02,748 INFO FIRST_STAGE
2023-09-20 09:36:02,748 INFO DIST_THRESHOLD : [0.2, 0.1, 0.1]
2023-09-20 09:36:02,748 INFO SECOND_STAGE
2023-09-20 09:36:02,748 INFO SCORE_THRESHOLD : [0.1, 0.1, 0.1]
2023-09-20 09:36:02,748 INFO POINT_THRESHOLD : [0, 0, 0]
2023-09-20 09:36:02,748 INFO DIST_THRESHOLD : [0.3, 0.15, 0.15]
2023-09-20 09:36:02,748 INFO TRACK_MERGE
2023-09-20 09:36:02,749 INFO ENABLE : True
2023-09-20 09:36:02,749 INFO CLASS_NAME : ['Vehicle', 'Pedestrian', 'Cyclist']
2023-09-20 09:36:02,749 INFO CLASS_THRESHOLD : [0.5, 0.4, 0.4]
2023-09-20 09:36:02,749 INFO REVERSE_TRACKING
2023-09-20 09:36:02,749 INFO ENABLE : True
2023-09-20 09:36:02,749 INFO POST_PROCESS
2023-09-20 09:36:02,749 INFO NAME : PostProcessor
2023-09-20 09:36:02,749 INFO CONFIG_LIST : [{'NAME': 'empty_track_delete', 'LEAST_AGE': 5, 'END_REMOVE': True}, {'NAME': 'velocity_optimize', 'HEADER_LENGTH': 3}, {'NAME': 'motion_classify'}, {'NAME': 'static_drift_eliminate'}]
2023-09-20 09:36:02,749 INFO REFINING
2023-09-20 09:36:02,749 INFO IOU_THRESHOLDS
2023-09-20 09:36:02,749 INFO Vehicle : 0.3
2023-09-20 09:36:02,749 INFO Pedestrian : 0.1
2023-09-20 09:36:02,749 INFO Cyclist : 0.1
2023-09-20 09:36:02,749 INFO ROOT_DIR : /detzero/tracking
2023-09-20 09:36:02,749 INFO CFG_FILE : cfgs/tk_model_cfgs/waymo_detzero_track.yaml
2023-09-20 09:36:02,749 INFO DATA_PATH : ../../data/waymo/waymo_infos_train.pkl
2023-09-20 09:36:02,749 INFO WORKERS : 4
2023-09-20 09:36:02,749 INFO BATCH_SIZE : 8
2023-09-20 09:36:02,749 INFO SPLIT : train
2023-09-20 09:36:02,749 INFO SAVE_LOG : False
2023-09-20 09:36:02,749 INFO -------------------------- Initialize WaymoTrackDataset --------------------------
2023-09-20 09:36:02,899 INFO ---------------------------- Running Tracking Module! ----------------------------
0%| | 0/1 [00:00<?, ?it/s]
0%| | 0/1 [00:06<?, ?it/s]
Traceback (most recent call last):
File "run_track.py", line 72, in
Kailthen
commented
1 year ago 同样的异常信息，您有进展吗
zorinasu
commented
11 months ago should use detection model result pkl path after --data_path
As mentioned in Intro of tracking, there is a script named "generate_gt_info.py" to generate gt_infos pickle files. However, I did not find that script in the project.