search

zanilzanzan / FuseNet_PyTorch

Joint scene classification and semantic segmentation with FuseNet
GNU General Public License v3.0
108 stars 34 forks source link

Output not normalized #11

Closed Oguked closed 6 years ago

Oguked commented 6 years ago

I think your model is missing a normalization operation like Softmax at the end, because the output values are big positiv/negativ numbers (-42, +30 etc.).

zanilzanzan commented 6 years ago

Hey there,

The cross-entropy loss function I used contains the softmax operation intrinsically. Please read the relevant documentation carefully before opening up an issue regarding functioning code.

Function details: https://pytorch.org/docs/stable/nn.html#cross-entropy Source: https://pytorch.org/docs/stable/_modules/torch/nn/functional.html#cross_entropy