search

sacmehta / ESPNet

ESPNet: Efficient Spatial Pyramid of Dilated Convolutions for Semantic Segmentation
https://sacmehta.github.io/ESPNet/
MIT License
538 stars 111 forks source link

We used the same code and procedure for the Camvid dataset. Please change the loadData.py file for CamVid dataset and you should be good to go #66

Closed pkuqgg closed 4 years ago

pkuqgg commented 5 years ago

We used the same code and procedure for the Camvid dataset. Please change the loadData.py file for CamVid dataset and you should be good to go

Originally posted by @sacmehta in https://github.com/sacmehta/ESPNet/issues/45#issuecomment-491573482

pkuqgg commented 5 years ago

Hello:

Thank you for reply!I know the way to train CamVid dataset. There is another important problem: from SegNet-Tutorial we can know CamVid dataset has 11 classes, but "pallete" show 12 classes. I try to change pallete in VisualizeResults, but the test result is bad.The last problem is how to test segmentation accuracy, I want to achieve this by fixing cityscapesscripts testing code, but still in working hardly.

I am really grateful for your reply, because my undergraduate degree major is statistics, my computer skills are relatively poor, and my classmates do not have the same research direction as me. I feel that I am in a big dilemma, thank you very much！I would be very grateful if you can provide the code for the test, so that I can learn from it.

Thanks!