sacmehta / ESPNet

ESPNet: Efficient Spatial Pyramid of Dilated Convolutions for Semantic Segmentation
https://sacmehta.github.io/ESPNet/
MIT License
541 stars 112 forks source link

There is a problem that ZeroDivisionError: float division by zero, how can I fix it? #27

Closed ghost closed 5 years ago

ghost commented 6 years ago

When I run the code by classes=2, I initial the code " model = net.ESPNet(classes=2)"

sacmehta commented 5 years ago

see #13

ghost commented 5 years ago

see #13

thanks for your help

Jee-King commented 5 years ago

Hi,could you give me some suggestions on classes=2.

ghost commented 5 years ago

Hi,could you give me some suggestions on classes=2.

you can initial your network with classes = 20, and add a conv layer to transpose 20 to 2 at the end.

Jee-King commented 5 years ago

Hi, How about the effect? I did not get good results by changing the model.py in #13. The smallest train loss I got was 0.036. Could you give me some suggestions? Thank you very much.