OrthoNets : Orthogonal Channel Attention Networks

PyTorch's implementation of the paper "OrthoNet : Orthogonal Channel Attention Networks".

Install

1. Install Linux
2. Install Anaconda
3. Install CUDA 11.X
4. Run the following comands in OrthoNet

   conda env create --file orthonet.yaml
   conda activate orthonet
   pip install --extra-index-url https://developer.download.nvidia.com/compute/redist nvidia-dali-cuda110

5. Edit settings.sh if necessary.

Dataset acquisition and installation

If you are installing these datasets for the first time, place them in OrthoNet/datasets. Otherwise, please edit settings.sh to set your dataset location.

Dataset Locations

ImageNet-1000

Visit https://www.image-net.org/ and download ImageNet 2014 Dataset.

MS-COCO

Train     Images      : http://images.cocodataset.org/zips/train2017.zip
Val       Images      : http://images.cocodataset.org/zips/val2017.zip
Train/Val Annotations : http://images.cocodataset.org/annotations/annotations_trainval2017.zip

Birds

Train/Val Images : https://www.kaggle.com/datasets/gpiosenka/100-bird-species/download?datasetVersionNumber=59

Places365

Train/Val Images : https://www.kaggle.com/datasets/benjaminkz/places365/download?datasetVersionNumber=1

Setting up the models

The models can be found at

https://drive.google.com/drive/folders/1N_L7Cy5I2lwVTnjjYan2Z0rXNVw8rN-c?usp=share_link

Download the "pretrained_models" folder and place it in OrthoNet. Typically, the files are downloaded in multiple zips. Verify all 14 models are present.

Training and Testing

Training and testing the models is done by the run.sh script. The specific commands are found below.

Testing Pretrained Models

Training Your Own Models

During training, FasterRCNN and MaskRCNN use the pretrained OrthoNet/FcaNet from the pretrained models folder.

Testing Your Trained Models