vcsl-owner

Thanks for your interest! We use 'you-get' with option '--format=flv360' to download the videos, and then use 'ffmpeg' to transcode.

Thanks for your interest. We re-run the scripts and are able to reproduce the numbers (R 45.10%, P 56.67%, F1 50.23%). The numbers in your "best-param-result" image is the same...

Due to the copyright limitation, we could not provide original video contents. But we have provided some other solutions for this problem (https://github.com/alipay/VCSL#solution-to-invalid-links). Please check it. Thanks!

Sorry that the training code of SPD will not be opened. But you can directly follow the training codes of YOLOv5.

Due to the copyright limitation, we could not provide original video contents. But we have provided some other solutions for this problem (https://github.com/alipay/VCSL#solution-to-invalid-links). Please check it. Thanks!

Sure. This is a nice direction for comparison between different frame features. Since we have open-sourced all the details of VCSL benchmark, we encourage you to directly compare between them...

Thank you for the feedback! For the "self-duplicate" pairs, we consider them as special and valid cases. The models are expected to consider one video as "copying" itself.   Our annotation...