Yunsheng Li

Haha, thanks for your confirmation. I will test my model again with the correct test set.

I have one more question. For the baseline result which is around ~15. I think you use bdd_peds_train.json file to train the model and use bdd_peds_val.json to test. However, I...

I cannot reproduce the result for SHTA either. When I try to unsupervised case, I can only get MAE of ~230 if I train the model only with the filtered...

What information do you need? I think both my paper and the Readme file contain all detail of training cyclegan.

opt.fineSize=1024 and opt.no_lsgan=False. I think when you said you have different amount of labels, it means the number of classes is different. In this case you should learn another segmentation...

I don't meet this problem.

Thanks for your interest in our work. But we do not have the pretrained ResNet-101 on hand.

I think I just use the pseudo labels as you described. But as you use a different DA framework, it's very hard for me to help you.

Thanks so much for your hard work. I do not realize the update of CycleGAN when I released the code. I will upload all my CycleGAN code later.

Please use the script I upload to implement MicroNet, instead of just using the default parameters.