search

mhamilton723 / STEGO

Unsupervised Semantic Segmentation by Distilling Feature Correspondences
MIT License
712 stars 143 forks source link

Training with custom dataset #46

Closed tanveer6715 closed 2 years ago

tanveer6715 commented 2 years ago

Hi,

Anyone here train STEGO using custom dataset? I tried to train the model with custom data but its accuracy and mIoU is "nan" as attached. Anyone who faced this problem please help me out. Thanks in advance.!!

ACC

tanveer6715 commented 2 years ago

Sorry I solved this problem so I am closing this issue.

gjackson1997 commented 2 years ago

Sorry I solved this problem so I am closing this issue.

Hey, im running into the same issue. How did you fix it? thanks

tanveer6715 commented 2 years ago

Sorry I solved this problem so I am closing this issue.

Hey, im running into the same issue. How did you fix it? thanks

Actually if you do not provide any labels so it cant compare validation prediction with the ground truth labels. If you want to check the mIoU or accuracy you need to provide ground truth labels for comparison.