Created by: yw2278

I'm trying to test the model on local videos, and seems there is no direct way to infer on the videos.