search

speedinghzl / pytorch-segmentation-toolbox

PyTorch Implementations for DeeplabV3 and PSPNet
MIT License
768 stars 167 forks source link

Confusion in performance #31

Closed forrestsz closed 5 years ago

forrestsz commented 5 years ago

hi,thanks for you nice work! I want to ask same question: In your implementation,what is your performance in cityscapes validation set and in what condition? I meet some confusion because i use your project only get 77.51 which batchsize is 12 in 4GPU and 77.40 which batchsize is 24 in 8GPU in val.but in the paper the mIoU can reach 78.5%. have you reach this result? thanks!

speedinghzl commented 5 years ago

The settings are listed in run_local.sh. The batch size is 8 on 4 GPUs. I think the settings are clear and easy to follow.

forrestsz commented 5 years ago

I have seen the run_local.sh, but i want to make sure that have you train model in pytorch 0.4 and close the ohem.thank!

speedinghzl commented 5 years ago

The answer is YES.