SSENet-pytorch icon indicating copy to clipboard operation
SSENet-pytorch copied to clipboard

Published 20 hours ago • verified publisher icon YudeWang

about the mIoU result

Open gengenkai opened this issue 4 years ago • 3 comments

I use your released weight (resnet38_aff_ser_0.3_424.pth) to perform the evaluation by your script (evalution.py). The mIoU result on the Pascal Voc dataset is 60.5. Could you tell me the reason? Thank you very much.

gengenkai avatar Apr 09 '20 03:04 gengenkai

hello，do you know the reason now? I got the miou result on the train set is 60.07

czzerone avatar Jun 19 '20 07:06 czzerone

@YudeWang hi, i got the same problem, could you give us some suggestion？

kkechen avatar Jul 15 '20 20:07 kkechen

The generated pseudo mask labels is evaluated on PASCAL VOC train set And please check the dataset to make sure the segmentation labels in train set come from original PASCAL VOC dataset instead of SBD, which is widely know as augmented VOC dataset.

YudeWang avatar Jul 21 '20 07:07 YudeWang