How to do the visualization of detection, tracking and prediction in the paper? Which script file should I use? Or how should I implement it? Looking forward to your reply, thank you!

[FunmiKesa]

Hi, the visualization codes are available here https://github.com/FunmiKesa/JLA/blob/master/src/lib/tracking_utils/visualization.py.

Let me know if that helps.

[manxunyu]

嗨，可视化代码可在此处获得https://github.com/FunmiKesa/JLA/blob/master/src/lib/tracking_utils/visualization.py。

让我知道这是否有帮助。

Thanks for your reply, but I still haven't finished the visualization. I checked the demo.py script in the FairMOT repo and the track_video.py script in your code is similar, but I run the script as follows to detect a video, and it cannot be detected: python src/track_video.py mot --load_model exp/mot/mot17_half_jla_ft_ch/model_30.pth --conf_thres 0.3 --input-video videos/MOT16-03.mp4 --forecast --gpus 0 The output information is as follows: Fix size testing. training chunk_sizes: [12] The output will be saved to /media/mxy/GD/dplearning/JLA/src/lib/../../exp/mot/default heads {'hm': 1, 'wh': 4, 'id': 128, 'reg': 2, 'fct': {'future_length': 60, 'past_length': 10, 'hidden_size': 512, 'input_size': 8, 'output_size': 4, 'fixed_length': False, 'use_embedding': False}} 2022-07-14 19:02:45 [INFO]: ['src/track_video.py', 'mot', '--load_model', 'exp/mot/mot17_half_jla_ft_ch/model_30.pth', '--conf_thres', '0.3', '--input-video', 'videos/MOT16-03.mp4', '--forecast', '--gpus', '0'] 2022-07-14 19:02:45 [INFO]: ['src/track_video.py', 'mot', '--load_model', 'exp/mot/mot17_half_jla_ft_ch/model_30.pth', '--conf_thres', '0.3', '--input-video', 'videos/MOT16-03.mp4', '--forecast', '--gpus', '0'] So I want to ask what's going on？

[FunmiKesa]

Hi, use demo.py instead of track_video.py. For example python demo.py mot --arch rnnforecast_34 --forecast --past_length 10 --future_length 35 --use_embedding --no_kf --load_model ../models/jla.pth --conf_thres 0.4 --input-video ../videos/MOT16-03.mp4 --output-root demos.