Demo on own video

[sophia-wright-blue]

Thank you for releasing this repo. Would it be possible to use this model to do a demo on my own video? What would the steps to do so be?

Thanks,

[fabienbaradel]

Hi @sophia-wright-blue, Thanks for your interest in our work. For sure you can run the code on your own video. The only one prerequisite is to first detect objects with mask-RCNN. One video should have an associated masks file. Once you have the detected objects and their masks it is straightforward using the validatefunction (cf below): https://github.com/fabienbaradel/object_level_visual_reasoning/blob/f504715c2ea9464e8dd4076584c30ab34937de22/inference/train_val.py#L148

[sophia-wright-blue]

thank you for the detailed reply @fabienbaradel , i'm gonna give that a shot, i look forward to understanding your model better, my eventual objective is to produce output similar to FIg 5 in your paper ("Object Level Visual Reasoning in Videos"), but with complete <human, verb, object> classification. Thanks again,