Framework for Development of a Multimodal Artificial Intelligence Based Forest Fire Prediction

This research proposes a multimodal Artificial Intelligence framework for forest fire prediction, combining sensor-driven meta-classification with advanced image analysis using YOLOv8. The system integrates real-time environmental data captured via a custom IoT device equipped with ESP32 and LoRaWiFi from fire-prone regions in southwestern Nigeria, alongside visual inputs from ground cameras and autonomous drones. Sensor data is processed through a layered ensemble of machine learning algorithms, while YOLOv8 identifies visual fire cues such as smoke and flames. A decision-level fusion mechanism merges both data streams to deliver high-accuracy predictions, minimize false alarms, and enable rapid, autonomous alerts. The framework demonstrates strong generalization, reliability, and scalability, making it a promising solution for low-resource, real-time forest fire monitoring and emergency response. Keywords: Framework, Development, Multimodal Artificial Intelligence, Forest Fire, Sensors Prediction,LoRaWiFi, YOLOV8 Image Analysis, Proceedings Citation Format Abiola O. A., Ajayi J.O & Akinola S.O. (2024): Framework for Development of a Multimodal Artificial Intelligence Based Forest Fire Prediction. Proceedings of the 37th iSTEAMS Multidisciplinary Bespoke Conference. 17th – 19th June, 2024. University of Ghana, Accra, Ghana. Pp 363-372. dx.doi.org