Trained on V-PRED. This has been trained for longer and has more training data compared to v1.0.
This was trained on more recent images. I will update it in the future.
