search

Tramac / awesome-semantic-segmentation-pytorch

Semantic Segmentation on PyTorch (include FCN, PSPNet, Deeplabv3, Deeplabv3+, DANet, DenseASPP, BiSeNet, EncNet, DUNet, ICNet, ENet, OCNet, CCNet, PSANet, CGNet, ESPNet, LEDNet, DFANet)
Apache License 2.0
2.85k stars 582 forks source link

mIoU is too low #96

Open xinanbuyi opened 4 years ago

xinanbuyi commented 4 years ago

  1. The oscillation of loss function is severe

  2. The segmentation effect is not good when using your own dataset, and the Miou is too low

Why, I look forward to your reply

Tramac commented 4 years ago

Hi, maybe the hyperparameters are not optimal or due to the reason of the dataset.

swjtulinxi commented 4 years ago

Hi，why is the miou lower than the paper，like pspnet，danet on cityscapes or voc2012，can we do a same results as the paper?

zzzzsh1 commented 2 years ago

i have the same problem

XXMxxm220 commented 1 year ago

i also