When I tested with the pre-trained checkpoint, I found that the L1 distance of the VoxCeleb and taichiHD datasets was better than the L1 distance in the paper. I want to know why.
Is the pre-trained checkpoint trained on the training set or on the entire dataset (including training set and testing set).
Thank you.
When I tested with the pre-trained checkpoint, I found that the L1 distance of the VoxCeleb and taichiHD datasets was better than the L1 distance in the paper. I want to know why. Is the pre-trained checkpoint trained on the training set or on the entire dataset (including training set and testing set). Thank you.