In many videos, the playback speed is significantly slowed down. This can be detrimental for models learning to recognize violence.
I wonder whether it would be possible to go back to original playback speed, without manually sifting through all the videos. I opened a random video and noticed quite some duplicate frames: could it be that I just have to remove the duplicate frames and stay at 30 FPS playback speed?
In the readme you mention the clips are sliced to 5 seconds / 30 fps, how was this done exactly? Was the source always 30 fps or higher? Or did some upsampling happen there as well?
In many videos, the playback speed is significantly slowed down. This can be detrimental for models learning to recognize violence.
I wonder whether it would be possible to go back to original playback speed, without manually sifting through all the videos. I opened a random video and noticed quite some duplicate frames: could it be that I just have to remove the duplicate frames and stay at 30 FPS playback speed?
In the readme you mention the clips are sliced to 5 seconds / 30 fps, how was this done exactly? Was the source always 30 fps or higher? Or did some upsampling happen there as well?