Multimodal Deep Transfer Learning with CNN-LSTM Fusion for Enhanced Forest Fire Detection and Risk Prediction

Forest fires are an escalating global threat, causing significant damage to ecosystems, wildlife and human settlements. The growing frequency and severity of wildfires, driven by factors like climate change and human activity, demand the development of advanced detection systems. Often slow and labor-intensive, traditional fire detection methods are insufficient for timely intervention. This paper explores machine learning techniques to improve wildfire detection across diverse datasets, including image and video data. Deep learning models, such as Transfer Learning using MobileNet, VGG16 and Xception, are assessed for image classification. Additionally, a video-based fire detection system is developed using EfficientNetB0 and LSTM networks to leverage, temporal analysis in video streams. This research paper highlights the potential of advanced neural networks to enhance wildfire detection systems, offering faster and more accurate responses to minimize the destructive impact of forest fires.