Test sets results

[arash-ashra]

For comparison in our paper, we are looking for the detailed test set results (class IoUs) of these prediction files that you shared: https://drive.google.com/drive/folders/156vMABydr7btdPDBU6b9J-e0jJHuPI73
Do you happen to have a snapshot of the submission results obtained with these predictions?
Thank you for your consideration.

[arash-ashra]

Is the best prediction files the same as the ones submitted to the first submissions in this snapshot? how about without SegFix results?

[PkuRainBow]

@arashash Yes, it is the same.

hrnet_w48_ocr_b_hrnet48_8_20000_xx_pretrain_freeze_bn_1_test_ms_6x_mIoU_842.zip is the result w/o SegFix, which achieves 84.2 on Cityscapes test set.

segfix_hrnet_w48_ocr_b_hrnet48_8_20000_xx_pretrain_freeze_bn_1_test_ms_6x_mIoU_845.zip is the result w/ SegFix, which achieves 84.5 on Cityscapes test set.

[arash-ashra]

Awesome. but there is no submission with 84.2 on Cityscapes test set right now. is it because it's replaced? if so do you have the screenshot of its detailed results?

[arash-ashra]

by the way, how should we cite these results? should we use this:
@Article{yuan2020segfix,
title={SegFix: Model-Agnostic Boundary Refinement for Segmentation},
author={Yuan Yuhui and Xie Jingyi and Chen Xilin and Wang Jingdong},
journal={arXiv preprint},
year={2020}
}
or your earlier works?

[PkuRainBow]

Thanks.

You might consider citing both OCR and SegFix as below (you can update the arxiv id of SegFix after the paper is released),

@Article{yuan2019ocr,
title={Object-Contextual Representations for Semantic Segmentation},
author={Yuan Yuhui and Chen Xilin and Wang Jingdong},
journal={arXiv preprint arXiv:1909.11065},
year={2019}
}

@Article{yuan2020segfix,
title={SegFix: Model-Agnostic Boundary Refinement for Segmentation},
author={Yuan Yuhui and Xie Jingyi and Chen Xilin and Wang Jingdong},
journal={arXiv preprint},
year={2020}
}

[PkuRainBow]

Here is the screenshot of our previous submission of HRNet + OCR (w/o SegFix). I hope it helps!

[arash-ashra]

Fantastic. Thank you so much