dustin-nguyen-qil / semi_vccreid Goto Github PK

Hello, your work is so interesting and it was attractive to me. But I have some questions now.
1.What's the difference between tsm_model_wo.pth.tar and tsm_model_w.pth.tar?
2.Could you share the pretrained model on CCVID?

Could you please upload the processed pickle files of VCCR?

Hello. In the section 4.2 "Implementation details" of your paper, you wrote that:

In testing stage, we applied the same sampling strategy on both datasets to form 8-frame input clips. The input clip is passed to the trained CNN backbone only, then the output frame-wise feature set is averaged as the video- wise person representation, which is then used to compute pair-wise similarities for matching stage.

Could you explain why in testing stage, only the CNN backbone is used but not other modules like TSM, texture branch, etc.?

Thank you.

I'm running "python datasets/prepare. Py - root"/media/dustin/DATA/Research/Video - -based ReID "-- dataset_name VCCR", The prompt is that the prepare. py file is not found.Could you please provide me with this file?