he010103
Question about low MOT benchmark result in training dataset
Hello, recently i was implementing your MTMC code on AIC track1 training data, However, even i followed the readme instruction and got the ' track1.txt ', then i seperated the track1.txt file by camera id in order to use AIC offered evaluation code for evluating, But the MOT benchmark result is extremely low. Therefore, i really hope that you can help me figure out why this happen!! Thanks a lot!!!!
Great, tomorrow i will visualize the results of each step and then send the screenshots to you! Thanks!
he010103 [email protected] 于2019年7月24日周三 下午7:58写道：
For better debugging, can you visualize the results of each step?
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/he010103/Traffic-Brain/issues/1?email_source=notifications&email_token=AI7KXN5UBB3BFBPPTLG4NU3QBA7VDA5CNFSM4IGMUIDKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD2WDGRA#issuecomment-514601796, or mute the thread https://github.com/notifications/unsubscribe-auth/AI7KXN7KNG3UBYXYZO2LSFDQBA7VDANCNFSM4IGMUIDA .
first step： filter single camera tracking result provided by AIC(deep sort && yolo v3
[image: trajectory_fusion.png] second step: trajectory fusion, cal average feature [image: filter_the_detection_file.png]
third step: clustering get mtmc result [image: clustering_for_seq1.png] but i did a little change in here for saving the tracking result which did not cross multi camera; [image: little_change_in_here_to_save_single_camera_result.png] last step: get track1.txt and then for evaluating the tracking result, i seperated the track1.txt according to the camera id;
result: [image: result.png]
qiuqi cao [email protected] 于2019年7月25日周四 下午11:55写道：
Great, tomorrow i will visualize the results of each step and then send the screenshots to you! Thanks!
he010103 [email protected] 于2019年7月24日周三 下午7:58写道：
For better debugging, can you visualize the results of each step?
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/he010103/Traffic-Brain/issues/1?email_source=notifications&email_token=AI7KXN5UBB3BFBPPTLG4NU3QBA7VDA5CNFSM4IGMUIDKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD2WDGRA#issuecomment-514601796, or mute the thread https://github.com/notifications/unsubscribe-auth/AI7KXN7KNG3UBYXYZO2LSFDQBA7VDANCNFSM4IGMUIDA .
in the first step, you should use my tracking results. in the last step, there is no need to seperated the track1.txt. You can draw the rectangle of the vehicle in the video for visualization. It will be better for you to find the problem.
Hi there,
I can't find 'images/1.jpg' when I run "cv_img = cv2.imread(opj(img_dir, s, c, 'images/1.jpg')) " in filter.py, how can I get it?
Please can you share the code folder with all settings
