remove mean embedding of dataset as preprocessing #52
Description
maybe this will slightly help overfitting to stock video distribution
need to find a way to efficiently take zero_masks in training and eval since the way I tried is super slow
{{ message }}