search

zhang-tao-whu / DVIS

DVIS: Decoupled Video Instance Segmentation Framework
MIT License
124 stars 7 forks source link

can not produce demos #17

Closed manli232111 closed 11 months ago

manli232111 commented 12 months ago

Hi authors,

Thanks for your great work! I try to produce some demo videos for VIPSeg dataset, but the segmentation results are poor, do I need to change some settings to reproduce the results shown in paper?

Thanks

manli232111 commented 12 months ago

Following the demo code, I got such demo, I am not sure how to fix it?

1
zhang-tao-whu commented 12 months ago

Please provide your running command. I guess you used the offline model but loaded the online weights, or used the online model but only loaded the segmenter weights.

manli232111 commented 12 months ago

Thanks for your reply!

I just git clone DVIS codebase and run the following code, I did not find any errors, could you pls help me check?

python demo_video/demo_long_video.py --config-file configs/youtubevis_2021/swin/DVIS_Offline_SwinL.yaml --output visual --input datasets/ytvis_2021/valid/JPEGImages/0e4068b53f/ --opt MODEL.WEIGHTS DVIS_offline_ytvis21_swinl.pth

zhang-tao-whu commented 11 months ago

Sorry for the delayed response. I downloaded the checkpoint and tested the visual results, and it works very well. download 00095

zhang-tao-whu commented 11 months ago

This is the complete visual output: demo_out-196.zip

zhang-tao-whu commented 11 months ago

I cannot speculate on what issue you are experiencing. If possible, please provide me with your operating environment, the command you are running, and a complete screenshot of the log.

manli232111 commented 11 months ago

The issue is due to the environment, I re-install the env and solved the bug. Thanks!