PyTorch for Semantic Segmentation

This repository contains some models for semantic segmentation and the pipeline of training and testing models, implemented in PyTorch.

Models

1. Vanilla FCN: FCN32, FCN16, FCN8, in the versions of VGG, ResNet and DenseNet respectively. (Fully convolutional networks for semantic segmentation)
2. U-Net (U-net: Convolutional networks for biomedical image segmentation)
3. SegNet (Segnet: A deep convolutional encoder-decoder architecture for image segmentation)
4. PSPNet (Pyramid scene parsing network)
5. GCN (Large Kernel Matters)
6. DUC, HDC (understanding convolution for semantic segmentation)
7. Deformable Convolution Network (in PSPNet version) (Deformable Convolutional Networks)

Visualization

Use powerful visualization of TensorBoard for PyTorch. Here to install.

Usage

1. Go to models directory and set the root path.
2. Go to datasets directory and do following the README.
3. Adjust the argument settings in train_psp.py (or train_fcn8.py, train_gcn.py) and run it.

Reference

I have borrowed some code from these nice repositories: [1], [2]. Thank them for the sharing.

TODO

1. DeepLab v3
2. RefineNet
3. CRFAsRNN
4. More dataset (e.g. ADE)