Bio-inspired colour opponency convolutional neural networks

---

1. Clone this Colour Opponency BioNet
2. Clone the CIFAR-10G for generalisation testset.
3. Clone the Ecoset If you encounter problems for download, you may try a direct download from CodeOcean's S3 bucket: "aws s3 cp --no-sign-request s3://codeocean-datasets/0ab003f4-ff2d-4de3-b4f8-b6e349c0e5e5/ecoset.zip ."
4. Extract categories as CIFAR-10, create Ecoset-10.(could use this script)

Expected directory structure

.
├── bionet
│   ├── Introduction.ipynb
│   ├── OpponencyCifarVGG16.ipynb
│   ├── OpponencyCifarResnet.ipynb
│   ├── OpponencyEcosetVGG16.ipynb
│   ├── OpponencyEcosetResnet.ipynb
│   ├── ResNet.py
│   └── VGG.py

├── data
│   ├── CIFAR-10G
│   ├── ecoset
│   └── ecoset-cifar10
├── logs
├── models
├── results
├── scripts
└── README.md

Introduction about Gabor filter with different parameters.

Click here Introduction of the gabor filter

Training and testing the model

The main script to handle training and testing are 4 .ipynb files. in the ./bionet dictionary. The general Colour Opponency GaborNet for VGG and Resnet are in file VGG.py and Resnet.py. The saved model could be found in dictionary models.