pytorch/vision

Add kinetics dataset to use new video reading API

Opened this issue · 0 comments

Currently, kinetics video dataset uses the old video reading API. Adding support for new video reading API would enable users to choose between CPU and GPU decoding functionality.