Multi Pathway 3D CNN Wit Conditional Random Field for Automated Segmentation of Multiple Sclerosis Lesions in MRI

Forest fires are a major environmental hazard that cause devastating damage to ecosystems, wildlife, and human communities. Early and accurate detection is crucial to minimize their impact. This project focuses on real-time forest fire detection using image classification techniques based on deep learning. Inspired by the base paper titled "Real-Time Detection of Forest Fires Using FireNet-CNN and Explainable AI Techniques", we aim to classify forest images into two categories: Fire and No Fire. The existing FireNet-CNN model serves as the baseline for performance comparison. To improve accuracy and computational efficiency, we propose the use of two advanced convolutional neural networks: EfficientNetV2 and MobileNetV3. These models are trained on a publicly available wildfire image dataset from Kaggle, which includes a balanced set of fire and non-fire images. We evaluate model performance using metrics such as accuracy, precision, recall, and F1-score. To enhance transparency and interpretability, Explainable AI (XAI) methods like Grad-CAM are applied to visualize areas of focus in the input images during prediction. The objective is to develop a lightweight, robust, and interpretable system for early wildfire detection, which can be integrated into surveillance and monitoring systems for rapid response and prevention.

Keywords:
Forest Fire Detection, FireNet-CNN, EfficientNetV2, MobileNetV3, Deep Learning, Image Classification, Explainable AI, Grad-CAM, Wildfire, Real-Time Monitoring.