AHMHCNN-mCBAM-for-Forest-Fire-Detection

HMHCNN-mCBAM: Hierarchical Multi-Head CNN with Modified CBAM for Visual Attention

This repository provides the implementation of HMHCNN-mCBAM, a novel deep learning architecture combining Hierarchical Multi-Headed Convolutional Networks with a Modified Convolutional Block Attention Module (mCBAM) for forest fire detection and classification. It is optimized for high-resolution aerial imagery tasks such as forest fire detection and damage assessment, delivering improved interpretability, accuracy, and computational efficiency.

---

🔍 Key Features

• 🧩 Hierarchical Convolution: Multi-scale feature encoding via grouped convolutions and resolution-aware pooling.
• 🧠 Multi-Head CNN Backbone: Parallel convolutional paths to learn diverse and rich spatial representations.
• 👁️‍🗨️ Modified CBAM (mCBAM): Extended attention mechanism with channel & spatial gating and normalization enhancements.

---

📁 Project Structure

File	Description
mcbam.py	Implements the Modified CBAM module with enhanced spatial and channel attention mechanisms.
ahmh.py	Defines the Adaptive Hierarchical Multi-Head Convolution layer for dynamic feature extraction.
HMHCNN-mCBAM.py	Main file for assembling and compiling the full HMHCNN-mCBAM model architecture.

---

🚀 Getting Started

✅ Prerequisites

• Python ≥ 3.8
• TensorFlow ≥ 2.8
• NumPy

📦 Install Dependencies

pip install tensorflow numpy

## 🗃 Dataset

This model is benchmarked on the **UAVs-FFDB** dataset:

**Citation**:  
Mowla, M. N., Asadi, D., Tekeoglu, K. N., Masum, S., & Rabie, K. (2024).  
*UAVs-FFDB: A high-resolution dataset for advancing forest fire detection and monitoring using unmanned aerial vehicles (UAVs).*  
Data in Brief, 55, 110706.  
[https://doi.org/10.1016/j.dib.2024.110706](https://doi.org/10.1016/j.dib.2024.110706)

---

## 📚 Citation

If you use this code or model in your research, please cite:

> M. N. Mowla, D. Asadi, S. Masum and K. Rabie,  
> "Adaptive Hierarchical Multi-Headed Convolutional Neural Network With Modified Convolutional Block Attention for Aerial Forest Fire Detection,"  
> *IEEE Access*, vol. 13, pp. 3412–3433, 2025.  
> doi: [10.1109/ACCESS.2024.3524320](https://doi.org/10.1109/ACCESS.2024.3524320)

### 📑 BibTeX
```bibtex
@article{mowla2025adaptive,
  title={Adaptive Hierarchical Multi-Headed Convolutional Neural Network With Modified Convolutional Block Attention for Aerial Forest Fire Detection},
  author={Mowla, M. N. and Asadi, D. and Masum, S. and Rabie, K.},
  journal={IEEE Access},
  volume={13},
  pages={3412--3433},
  year={2025},
  doi={10.1109/ACCESS.2024.3524320}
}

## 📄 License

This project is licensed under the **Creative Commons Attribution 4.0 International (CC BY 4.0)** License.  
You are free to share and adapt this work, even for commercial use, as long as you provide appropriate credit.  
For more details, visit: [https://creativecommons.org/licenses/by/4.0/](https://creativecommons.org/licenses/by/4.0/)