Open ZeroCool22 opened 6 years ago
Can you show me the result video clip containing [source, transformed, mask]?
How many training data did you trained on? It seems like the training data is lack of diversity. The model can not learn proper masks.
Here is the video you ask me for: https://www.youtube.com/watch?v=prdMPR7IKsg
OUTPUT_VIDEO:
https://www.youtube.com/watch?v=wew_q9yJMug
Train preview images:
Also there doesn't seems to be any attempt for try to match colors?