ORN-pytorch

Baseline code for [Oriented Response Networks]

Please follow the instruction below to setup it and run the experiment demo.

Prerequisites

• Linux (Ubuntu 16.04LTS)
• NVIDIA-1080ti + CUDA-9.0 + CuDNN7.0
• Pytorch-1.0.0 + Torchvision

Getting started

1. install the dependencies (required by the demo code):

   • pillow: pip install Pillow

2. clone the this branch:

   # git version must be greater than 1.9.10
   git clone https://github.com/SDL-GuoZhao/ORN-master

3. setup orn:

   cd ORN-pytorch
   export DIR=$(pwd)
   python setup.py develop

4. run the MNIST-Variants demo:

   cd $DIR
   mkdir outputs
   # train baseline CNN on MNIST
   python main.py --net cnn --dataset MNIST
   # train ORN on MNIST
   python main.py --net orn --dataset MNIST
   # train baseline CNN on MNIST-rot
   python main.py --net cnn --dataset MNIST-rot
   # train ORN on MNIST-rot
   python main.py --net orn --dataset MNIST-rot

Evaluate model

```bash
cd $DIR
# test model on MNIST
python utils/eval_model.py --model-path outputs/path_to_your_ckpt.pth --dataset MNIST
```

Citation

If you use the code in your research, please cite:

@INPROCEEDINGS{Zhou2017ORN,
    author = {Zhou, Yanzhao and Ye, Qixiang and Qiu, Qiang and Jiao, Jianbin},
    title = {Oriented Response Networks},
    booktitle = {CVPR},
    year = {2017}
}