search

UkcheolShin / CRM_RGBTSeg

Official implementation of the paper "Complementary Random Masking for RGB-T Semantic Segmentation."
MIT License
52 stars 12 forks source link

Why is the result of online evaluation different from testing? #9

Open Stark320 opened 3 months ago

Stark320 commented 3 months ago

I use the same data_loader and the same model, the only difference is that using multiple GPUs during online evaluation and single GPU during testing. Could you explain this? Thanks a lot. Great work by the way.

HaotianLi816 commented 3 months ago

Same problem, have you solved it? I found that the IoU in tensorboard is higher than the result in "test.py".