Emergency Reporting Platform AI-Driven Fire Incident Management and Response

Abstract

Forests are vital natural resources that provide numerous benefits to humanity, yet they are increasingly threatened by natural disasters such as forest fires. These fires contribute significantly to global warming and pose a risk to ecosystems and life on Earth. Early detection of forest fires is crucial for timely response and mitigation efforts. This study explores the use of artificial intelligence (AI)-based computer vision techniques for automatic fire and smoke detection from images. Convolutional Neural Networks (CNNs) have demonstrated exceptional performance in image classification tasks, but their training times are often prohibitive, and pre-trained models can underperform when applied to limited datasets. To address these challenges, we employ transfer learning on pre-trained CNN models, optimizing their performance on new datasets. Additionally, we integrate a technique called Learning without Forgetting (LwF), which allows the model to learn new tasks, such as fire detection, while retaining its classification abilities from previous tasks. This approach ensures efficient and accurate fire detection without compromising the model's performance on original datasets. The system is implemented using a technology stack that includes Python, Django, MySQL, HTML, CSS, JS, and Bootstrap to create a robust web-based platform for real-time fire detection.