Thank you so much for providing the code for your paper :). Any chance I could get my hands on the checkpoints, and pre-trained models?
I also wanted to clarify some things regarding training the model. The training is done on unlabeled images first, where the segmentation branch is completely ignored. After that, you use the labeled dataset to refine the generator. Would it be possible to elaborate on that please? The supplementary material covers the cases of training the generator and the encoder. Nothing is mentioned about training the generator with labeled and unlabeled datasets.
Hi there!
Thank you so much for providing the code for your paper :). Any chance I could get my hands on the checkpoints, and pre-trained models?
I also wanted to clarify some things regarding training the model. The training is done on unlabeled images first, where the segmentation branch is completely ignored. After that, you use the labeled dataset to refine the generator. Would it be possible to elaborate on that please? The supplementary material covers the cases of training the generator and the encoder. Nothing is mentioned about training the generator with labeled and unlabeled datasets.