Repository files navigation

[Home][Project][Paper][Supp][Poster]

• Tested with PyTorch 1.12.0 (Ubuntu / GTX 2080 Ti)
• A New helper functionupgrade_to_orn for easy model conversion.
• Predefined ORN-upgraded models (OR-VGG, OR-ResNet, OR-Inception, OR-WRN, etc.).

Please check thepytorch-v2 branch for more details.

Thetorch branch contains:

• the officialtorch implementation of ORN.
• theMNIST-Variants demo.

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

• Linux (tested on ubuntu 14.04LTS)
• NVIDIA GPU + CUDA CuDNN (CPU mode and CUDA without CuDNN mode are also available but significantly slower)
• Torch7

You can setup everything via a single commandwget -O - https://git.io/vHCMI | bashor do it manually in case something goes wrong:

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

   • torchnet:luarocks install torchnet
   • optnet:luarocks install optnet

2. clone the torch branch:

   # git version must be greater than 1.9.10git clone https://github.com/ZhouYanzhao/ORN.git -b torch --single-branch ORN.torchcd ORN.torchexport DIR=$(pwd)

3. install ORN:

   cd$DIR/install# install the CPU/GPU/CuDNN version ORN.bash install.sh

4. unzip the MNIST dataset:

   cd$DIR/demo/datasetsunzip MNIST

5. run the MNIST-Variants demo:

   cd$DIR/demo# you can modify the script to test different hyper-parametersbash ./scripts/Train_MNIST.sh

If you run into'cudnn.find' not found, update Torch7 to the latest version viacd <TORCH_DIR> && bash ./update.sh then re-install everything.

CIFAR 10/100

You can train theOR-WideResNet model (converted from WideResNet by simply replacing Conv layers with ORConv layers) on CIFAR dataset withWRN.

dataset=cifar10_original.t7 model=or-wrn widen_factor=4 depth=40 ./scripts/train_cifar.sh

With exactly the same settings, ORN-augmented WideResNet achieves state-of-the-art result while using significantly fewer parameters.

Table.1 Test error (%) on CIFAR10/100 dataset with flip/translation augmentation)

ImageNet

The effectiveness of ORN is further verified on large scale data. The OR-ResNet-18 model upgraded fromResNet-18 yields significant better performance when using similar parameters.

Table.2 Validation error (%) on ILSVRC-2012 dataset.

You can usefacebook.resnet.torch to train theOR-ResNet-18 model from scratch or finetune it on your data by using thepre-trained weights.

-- To fill the model with the pre-trained weights:model=require('or-resnet.lua')({tensorType='torch.CudaTensor',pretrained='or-resnet18_weights.t7'})

A more specific demo notebook of using the pre-trained OR-ResNet to classify images can be foundhere.

Please check thepytorch-v2 branch for more details.

Thepytorch branch contains:

• the officialpytorch implementation of ORN(alpha version supports 1x1/3x3 ARFs with 4/8 orientation channels only).
• theMNIST-Variants demo.

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

• Linux (tested on ubuntu 14.04LTS)
• NVIDIA GPU + CUDA CuDNN (CPU mode and CUDA without CuDNN mode are also available but significantly slower)
• PyTorch

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

   • tqdm:pip install tqdm
   • pillow:pip install Pillow

2. clone the pytorch branch:

   # git version must be greater than 1.9.10git clone https://github.com/ZhouYanzhao/ORN.git -b pytorch --single-branch ORN.pytorchcd ORN.pytorchexport DIR=$(pwd)

3. install ORN:

   cd$DIR/installbash install.sh

4. run the MNIST-Variants demo:

   cd$DIR/demo# train ORN on MNIST-rotpython main.py --use-arf# train baseline CNNpython main.py

Thecaffe branch contains:

• the officialcaffe implementation of ORN(alpha version supports 1x1/3x3 ARFs with 4/8 orientation channels only).
• theMNIST-Variants demo.

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

• Linux (tested on ubuntu 14.04LTS)
• NVIDIA GPU + CUDA CuDNN (CPU mode and CUDA without CuDNN mode are also available but significantly slower)
• Caffe

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

   • idx2numpy:pip install idx2numpy

2. clone the caffe branch:

   # git version must be greater than 1.9.10git clone https://github.com/ZhouYanzhao/ORN.git -b caffe --single-branch ORN.caffecd ORN.caffeexport DIR=$(pwd)

3. install ORN:

   # modify Makefile.config first# compile ORN.caffemake clean&& make -j"$(nproc)" all

4. run the MNIST-Variants demo:

   cd$DIR/examples/mnistbash get_mnist.sh# train ORN & CNN on MNIST-rotbash train.sh

Due to implementation differences,

• upgrading Conv layers to ORConv layers can be done by adding anorn_param
• num_output of ORConv layers should be multipied by nOrientation of ARFs

Example:

layer {type:"Convolution"name:"ORConv"bottom: "Data" top: "ORConv"# add this line to replace regular filters with ARFsorn_param {orientations:8}param { lr_mult:1 decay_mult: 2}convolution_param {# this means 10 ARF feature mapsnum_output:80kernel_size:3stride:1pad:0weight_filler { type:"msra"}bias_filler { type:"constant"value: 0}}}

Check the MNIST demoprototxt (and itsvisualization) for more details.

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}}