SoccerNet / sn-gamestate

[CVPRW'24] SoccerNet Game State Reconstruction: End-to-End Athlete Tracking and Identification on a Minimap (CVPR24 - CVSports workshop)
228 stars 45 forks source link

GT file not found eval/gt/valid/SNGS-021.txt #19

Open yijietan opened 1 week ago

yijietan commented 1 week ago

I'm running the baseline on the single video (default is SNGS-021), and after the visualization video is created and when tracking is done, an error occurs, saying that the GT file cannot be found.

Am I missing any special configs in soccernet.yaml or somewhere else? As far as I know only my pitch and calibration fields are nbjw_calib instead of tvcalib

LianyouII commented 1 week ago

The same issue also troubled me. You need to modify the "tracklab/tracklab/configs/dataset/soccernet_gs.yaml" file, changing the "GT_LOC_FORMAT" to '{gt_folder}/{seq}/Labels-GameState.json'. The complete file is as follows:

截图 2024-10-12 23-52-22

Additionally, in my case, I also needed to modify the "trackeval" library in the environment, specifically the "your_env_path.../site-packages/trackeval/datasets/soccernet_gs.py" file, changing self.benchmark ='SoccerNetGameState'.

yijietan commented 1 week ago

Thanks, that works!

I suppose then without a GT file, like for those in the challenge dataset or any other video datasets that you input, is there a way to turn off the evaluation step?

LianyouII commented 1 week ago

Thanks, that works!

I suppose then without a GT file, like for those in the challenge dataset or any other video datasets that you input, is there a way to turn off the evaluation step?

Carefully read the soccernet.yaml file comments and set nframes not equal to -1, which will skip evaluation step.

yijietan commented 1 week ago

Thanks, that works! I suppose then without a GT file, like for those in the challenge dataset or any other video datasets that you input, is there a way to turn off the evaluation step?

Carefully read the soccernet.yaml file comments and set nframes not equal to -1, which will skip evaluation step.

Thanks for pointing out, just saw that comment shortly before you posted the reply.