wkentaro / pytorch-fcn

PyTorch Implementation of Fully Convolutional Networks. (Training code to reproduce the original result is available.)
MIT License
1.74k stars 480 forks source link
computer-vision convolutional-networks deep-learning fcn fcn8s pytorch semantic-segmentation

pytorch-fcn

PyPI Version Python Versions GitHub Actions

PyTorch implementation of Fully Convolutional Networks.

Requirements

Installation

git clone https://github.com/wkentaro/pytorch-fcn.git
cd pytorch-fcn
pip install .

# or

pip install torchfcn

Training

See VOC example.

Accuracy

At 10fdec9.

Model Implementation epoch iteration Mean IU Pretrained Model
FCN32s Original - - 63.63 Download
FCN32s Ours 11 96000 62.84
FCN16s Original - - 65.01 Download
FCN16s Ours 11 96000 64.91
FCN8s Original - - 65.51 Download
FCN8s Ours 7 60000 65.49
FCN8sAtOnce Original - - 65.40
FCN8sAtOnce Ours 11 96000 64.74

Visualization of validation result of FCN8s.

Cite This Project

If you use this project in your research or wish to refer to the baseline results published in the README, please use the following BibTeX entry.

@misc{pytorch-fcn2017,
  author =       {Ketaro Wada},
  title =        {{pytorch-fcn: PyTorch Implementation of Fully Convolutional Networks}},
  howpublished = {\url{https://github.com/wkentaro/pytorch-fcn}},
  year =         {2017}
}