The final model. Trained on 768x768 images, so it better works with larger resolutions than previous ones.
The final model. Trained on 768x768 images, so it better works with larger resolutions than previous ones.