is there any way to inference in a long time video

[come105]

When doing inference in a video, you need to use
inference_state = predictor.init_state()
to initialize the state, which will load all frames of the video, and if the video is large, it will take up a lot of memory or even fail.
I don't know why it needs to load all frames, is there any way to use segmentation or other methods to do inference in a large video?

[ovalerio]

Hi @come105,

I was also having also issues when using a large video. You can try using smaller video chunks and feed that to the model. Another option is to offload the video to the CPU. I understand that takes a bit longer but it is a good workaround if you are running out of GPU memory.

inference_state = predictor.init_state(video_path=video_dir, offload_video_to_cpu=True)

The init_state method has other few flags that you can play with.

[heyoeyo]

It's possible to avoid loading the frames into memory, but requires some (small) code changes. With some other changes (to avoid caching results per-frame), you can keep the VRAM use under 2GB for any video length. There's more info in issue #264.