How to reproduce the SoTA result of Cityscapes?

[JasiRose]

Hello, thanks for your impressive work! However, I am not able to reproduce the results of Cityscapes using this repo. In the paper, the mIOU of HRNetv2 increased to 81.4, and the mIOU of HRNetv2+OCR increased to 83.2 on the Cityscapes dataset. When I ran the script "run_h_48_d_4_contrast_mem.sh", the mIOU is hard to converge. Also I notice that the script of HRNet+OCR+Contrastive based on memory bank is not given, could you suggest me how to achieve the experimental results in the paper?

[tfzhou]

Hi @JasiRose Thanks for your interests. I would suggest you to provide more details about your results. If you'd like to reproduce the results in the paper, please change the number of training iterations as mentioned in Sec.4.1 in our paper. In addition, I'd like to clarify that we release our algorithm on openseg.pytorch with pytorch-1.7.1 for easier use, but the experiments in the paper are ran on openseg.pytorch with pytorch-0.4.1. There are some performance difference between the two codebases, especially for deeplabv3. For the script for OCR, I assume that it will be easy to write based on the provided scripts, but if you still have difficulties, please come back to me I will guide.