Dual Attention for Image Classification

Introduction

This project combine Spatial attention from the BAM model with Frequency Channel Attention from FcaNet network to improve CNN model for medical image classification.

Use Spatial attention in BAM:

Frequency Channel Attention:

Dependency

• python3.10
• tensorflow2.15
• The code is work on Ubuntu 20.04.6 LTS

Recommended

• Ubuntu 20.04 with tensorflow GPU edition
• Kaggle or Colab free version.
• I implement code on Colab Notebook.

Getting Started

git https://github.com/nguyen599/Dual-attention-cnn-for-image-classification.git

Dataset

• ISIC 2018 task 3 - Skin lesion diagnosis dataset. There are 10,015 images in this dataset with seven different categories.

• KvasirPogorelov dataset - Dataset includes 4000 endoscopic gastrointestinal diseases and comprises eight classes, each containing 500 images.

You can use my proccessed data. Your dataset structure should like:

./data
├── train
│   ├── class 1
│   │   ├── image 1
│   │   ├── image 2
│   │   └── ...
│   ├── class 2
│   └── ...
└── test
    ├── class 1
    │   ├── image 1
    │   ├── image 2
    │   └── ...
    ├── class 2
    └── ...