Thank you for open sourcing your paper! The results presented in the paper are extremely compelling.

I have some questions and requests:

1. do you plan to release all of the configs you used in the paper? I would greatly appreciate it so I can reproduce the Cityscapes results. Also, are there any weights that you can share to verify evaluations for Cityscapes and Pascal VOC?
2. do you have a specific way to evaluate the trained models? I was going to start training the only available config ('cps_meanteacher_3b_w1.5_...'), and was wondering if the saved checkpoints can be directly used for tools/test.py.
3. For ExpSemiLossCPSFAWS3 (which I assume is the Lst and Lss losses) seem to not use knowledge selection. Reading the code, the self.prob seems to be the threshold, but I don't see anywhere where that is set to 0.95 as in the paper.

Welcome update to OpenMMLab 2.0

I am Vansin, the technical operator of OpenMMLab. In September of last year, we announced the release of OpenMMLab 2.0 at the World Artificial Intelligence Conference in Shanghai. We invite you to upgrade your algorithm library to OpenMMLab 2.0 using MMEngine, which can be used for both research and commercial purposes. If you have any questions, please feel free to join us on the OpenMMLab Discord at https://discord.gg/A9dCpjHPfE or add me on WeChat (ID: van-sin) and I will invite you to the OpenMMLab WeChat group.

Here are the OpenMMLab 2.0 repos branches:

Attention: please create a new virtual environment for OpenMMLab 2.0.

Hi! About the Table 1 in your paper I have two questions: (1) What's the meaning of "introduce the unlabeled dataset with a total of 10582 images" in the notion of Table1? What is the difference between two settings? (2)What exactly is the "enhanced training scheme", I haven't found any related illustration about it in Table 4, as said in the notion of Table 1? Looking forward to your reply. Thanks a lot!

in "config/semi_ablations/cps_meanteacher_3b_w1.5_w.1.0_FDmt1.5.py"
you set annotation of train_unsup as 'SegmentationClassAug'.
What is mean about 'SegmentationClassAug?

As I know, in semi-supervised learning, actually don't use annotation of unlabeled data.

How do I change my own data set? Do I just need to change the number of categories and the routing of the data set j? This calculates the IOUs for all categories except background to be 0