same classification result with every pixel

hellochick / ICNet-tensorflow

TensorFlow-based implementation of "ICNet for Real-Time Semantic Segmentation on High-Resolution Images".

406 stars 153 forks source link

Open AishuaiYao opened 4 years ago

AishuaiYao commented 4 years ago

When inference, why is my classification result of each pixel the same, all is No.5 classes？

AishuaiYao commented 4 years ago

I found the cause of the error is bn operation, but after the modification, the model effect is very bad, how about you