Pytorch Fully Convolutional Network

Paper implementation of Fully Convolutional Networks for Semantic Segmentation.

Requirements

• Python 3.6
• PyTorch

Installation

git clone https://github.com/andersskog/pytorch-fully-convolutional-network.git
cd pytorch-fully-convolutional-network
pip3 install -r requirements.txt

Note: for MacOS, see http://pytorch.org/ for other OS.

Dataset

Images and annotations are a subset of ADE20K Dataset. This dataset is build to predict a total of 150 classes.

Usage

python3 main.py train_list val_list dataset_path

Example

python3 main.py train.txt val.txt data