Genie-Kim
What is the aux_gts in the training code? (Dataloader)
In your meta-learning algorithm in your train-code, What does aux_gts exactly mean? Is the data gts and aux_gts are the same?
Thank you for providing the code of your work!
Thank you for your interest!
We adopt the auxiliary per-pixel cross-entropy loss proposed in PSPNet [1] with a coefficient of 0.4 to make a fair comparison with the RobustNet [2]. The aux_gts are the same as the original gts in our settings. However, in RobustNet [2], it seems that aux_gts are different depending on their ablation setting. We just kept that option as is.
[1] Zhao et al, Pyramid scene parsing network. In CVPR, 2017. [2] Choi et al, Robustnet: Improving domain generalization in urban-scene segmentation via instance selective whitening. In CVPR, 2021.
