Reproduce Tab. 2 results

[perladoubinsky]

Hello,

Thank you for the nice work !

I was able to reproduce the results of Tab. 1 but I am having trouble for Tab. 2 (especially line 3 and 4). Could specify the settings you used for training DAFormer on Cityscapes?

In addition, for the results DAFormer + DGInStyle (line 4), how much synthetic data did you use (6000 images or same as the number of Cityscapes images)? The data is taken from the same generated dataset as for Tab. 1 results?

Thanks!

[yurujaja]

Hello,

For Tab. 2, the generated data differs from the dataset generated in Table 1. In this case, Cityscapes is used as the source domain data for the whole pipeline.

[perladoubinsky]

Ok I see, thank you. For training DAFormer on Cityscapes without the synthetic data, can you specify the settings that differ from GTA training (rare class sampling, shade etc). I cannot reproduce the results of Tab.2 line 3.

Thank you!

[bangbang-bangbang]

Hello,

Thank you for the nice work !

I was able to reproduce the results of Tab. 1 but I am having trouble for Tab. 2 (especially line 3 and 4). Could specify the settings you used for training DAFormer on Cityscapes?

In addition, for the results DAFormer + DGInStyle (line 4), how much synthetic data did you use (6000 images or same as the number of Cityscapes images)? The data is taken from the same generated dataset as for Tab. 1 results?

Thanks!

Hello @perladoubinsky ,my result for training HRDA on Cityscapes without the synthetic data is 49~50mIoU, How about yours?

[ywher]

I think there are quit a lot of bugs in this code, making it hard to reproduce the result.....
For example, the class DGDataset lack the input "source2" in build.py.....