some questions about the classifier model of deeplab

wasidennis / AdaptSegNet

Learning to Adapt Structured Output Space for Semantic Segmentation, CVPR 2018 (spotlight)

850 stars 203 forks source link

some questions about the classifier model of deeplab #20

Closed zhyx12 closed 6 years ago

zhyx12 commented 6 years ago

the return statement in forward function of Classifier_Module looks like misaligned.

https://github.com/wasidennis/AdaptSegNet/blob/e47871be396a08fb10f608889ba85620312d2d24/model/deeplab_multi.py#L114-L118

https://github.com/wasidennis/AdaptSegNet/blob/e47871be396a08fb10f608889ba85620312d2d24/model/deeplab_vgg.py#L17-L21

wasidennis commented 6 years ago

Thank you for noticing. This is a bug also later on fixed by the reference repo.

The bug is essentially making our model have only 2 effective ASPP layer, and we don't think it will affect the performance much. We will try to re-train the model based on the fixed bug.

zhyx12 commented 6 years ago

Got it. Thanks for your explanation.