davidtvs / PyTorch-ENet

PyTorch implementation of ENet
MIT License
389 stars 129 forks source link

Wrong about test #33

Closed ChaosNN closed 4 years ago

ChaosNN commented 4 years ago

The model I train and you provide both have this problem.

Avg. loss: 0.0000 | Mean IoU: nan unlabeled: nan road: nan sidewalk: nan building: nan wall: nan fence: nan pole: nan traffic_light: nan traffic_sign: nan vegetation: nan terrain: nan sky: nan person: nan rider: nan car: nan truck: nan bus: nan train: nan motorcycle: nan bicycle: nan

davidtvs commented 4 years ago

if you ran cityscapes in test mode that's expected since the test set of cityscapes is private; therefore, there are no images

MaximumProgrammer commented 4 years ago

So how to test on another test set after training ?

davidtvs commented 4 years ago

You can test on CamVid, but for other datasets you have to do your own dataloader and the test script (maybe reusing the test function and/or test class).

MaximumProgrammer commented 4 years ago

Ok thank you, kind regards.