Hi,
Thanks a lot for this repo,
I'm trying to use OpenCV instead of ffmpeg in visual_compute_embs function.
When trying to process the same video with same texts mostly ffmpeg's scores higher than OpenCV ones.
From the visual debugging I did (saving the numpy arrays as frames) I saw that OpenCV preserves the color but doesn't preserve in ffmpeg. And still, the evaluation is better.
Any ideas how I can improve OpenCV's performance? And why this is the case?