daipengwa
commented
1 year ago Sorry for late reply. In our case, the consistency of inferred test video is improved while using the consistency regularization. " If the temporal consistency of the test video is not similar to that of the training video", I am not sure what you mean? The videos used for training and testing are different in our case.
Hello, you need ground truth video to calculate the loss of temporal consistency. If the temporal consistency of the test video is not similar to that of the training video, will the consistency of the inferred test video be poor？