Random question … do you know what would be the best way to support batched dataloader with multiple workers (in order to support costly image transformations) and also be able to stream in long videos?
I can just use cv2.videocapture to capture raw frames very fast, but can’t embed this into dataset with multiple workers I don’t believe ( not pickleable).
Random question … do you know what would be the best way to support batched dataloader with multiple workers (in order to support costly image transformations) and also be able to stream in long videos?
I can just use cv2.videocapture to capture raw frames very fast, but can’t embed this into dataset with multiple workers I don’t believe ( not pickleable).