search

speedinghzl / pytorch-segmentation-toolbox

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

The activation "ReLU" in models? #20

Closed CS-Jackson closed 5 years ago

CS-Jackson commented 5 years ago

Hi, @speedinghzl . Thanks for the great job. I have a little problem about the activation using in the model. As in the inplaceABN paper, the ReLU activation is not invertible. Do you use the ReLU activation to train the model?

speedinghzl commented 5 years ago

Yes, I use the ReLU activation without in-place operation.

CS-Jackson commented 5 years ago

Yes, I use the ReLU activation without in-place operation.

Thank you for the answer. By the way, can you share the weight training on CityScapes?

speedinghzl commented 5 years ago

I will release the trained model later. Now I am focusing on the conference submission. If you want to obtain the high-performance segmentation result of Cityscapes, I refer you to my another repo CCNet.

erichhhhho commented 5 years ago

@speedinghzl Any update for the pretrained PSPNet?