Firstly, many thanks for making the codebase public.
Now, I wanted to make a webcam based demo (creating a buffer of latest N video frames and recognizing actions for them) with models trained on something-something-v2 datasets with 'RGB', 'PA' modality. I am not able to find out a way to do the same using the existing code... :( Please help!
Firstly, many thanks for making the codebase public. Now, I wanted to make a webcam based demo (creating a buffer of latest N video frames and recognizing actions for them) with models trained on something-something-v2 datasets with 'RGB', 'PA' modality. I am not able to find out a way to do the same using the existing code... :( Please help!