The author mentions in the README that the training details can be found in test_dev.py.
However, test_dev.py is missing many of the methods suggested in the paper.
For example, depth consistency loss, feature embedding, and the options used for training (e.g., occlusion mask, cutmix ...) are missing.
This seems to be information that is hard to find in test_dev.py, and I wonder if there are any plans to follow up on this.
Has anyone been able to reproduce it?
The author mentions in the README that the training details can be found in test_dev.py.
However, test_dev.py is missing many of the methods suggested in the paper.
For example, depth consistency loss, feature embedding, and the options used for training (e.g., occlusion mask, cutmix ...) are missing.
This seems to be information that is hard to find in test_dev.py, and I wonder if there are any plans to follow up on this.
Has anyone been able to reproduce it?