Artificial Intelligence for Early Forest Fire Detection and Smart Afforestation
Loading...
Files
Date
Journal Title
Journal ISSN
Volume Title
Publisher
Unuversity Setif 1 Ferhat Abbas . Faculty of Sciences
Abstract
This thesis presents an intelligent system for early forest fire detection and smart afforestation in Algerian ecosystems using two complementary deep learning pipelines. Approach 1 combines a fine-tuned ResNet50 classifier with YOLOv11s, achieving 99.55% classification accuracy, a mAP@0.5 of 44.22%, and 10.3 FPS. Approach 2 adopts a unified YOLOv11 architecture for both classification and detection, achieving 99.38% accuracy, a higher mAP@0.5 of 76.86%, and 21.1 FPS, making it the most suitable solution for real-time deployment. Both approaches integrate a confidence-based fire intensity module with three severity levels (Strong, Medium, and Weak) and a smart afforestation recommendation engine based on the hybrid MAX+MEAN decision rule. The complete system is deployed as an interactive Streamlit dashboard for fire monitoring and post-fire ecological restoration.
