How to train a tlt classification model based on video inputs

Hi everyone,
I’m using tlt-streamanalytics:v2.0_py3 .
I have a video dataset and I would like to know if I can train a tlt classification model based on video inputs, instead of image inputs. If it is possible in tlt, then which source code should I modify?

Video input is not supported in tlt classification training. You can use ffmpeg to generate images for training.

1 Like

thanks a lot, @Morganh.
the problem is I have a large video dataset, extract frames from them need lots of memory and time-consuming. I hope that in the future this option will be supported.