FireTools: Integrated Tools for Enhanced Wildfire Detection and Segmentation

November 28, 2025

Wildfires represent a serious threat to human lives, ecosystems, and infrastructure, particularly in Portugal, where severe fire events have become increasingly frequent. Current wildfire detection approaches rely heavily on manual video analysis, a process that is slow and prone to error. Although automated computer vision methods show strong potential, they remain limited by scarce training data, class imbalance, and challenges in detecting small or partially obscured fires. This thesis introduces FireTools, a modular system for wildfire detection and segmentation using RGB. The system comprises two main components: FireDetect, which enables fast detection of fire and smoke through bounding box object detection, and FireSeg, which performs pixel-level segmentation. These components are integrated into a web-based platform that supports GPU acceleration, video upload, and synchronized visualization of detection and segmentation results using temporal timelines, enabling efficient use in real-world firefighting and emergency response scenarios. The system is validated through three experimental studies. The first evaluates the impact of dataset scaling and data diversity on detection performance. The second compares YOLO and Faster R-CNN, analyzing trade-offs between inference speed and detection accuracy. The third assesses segmentation architectures, including DeepLabv3+ and EfficientSeg. Results show that dataset diversity contributes more to detection performance than dataset size alone. YOLO achieves 5.51 times faster inference, while Faster R-CNN provides higher detection accuracy. For segmentation, a hybrid DeepLabv3+ YOLO Smoke approach achieves the best overall performance, improving smoke IoU by 10.8%.

Authors in the community:

• 

  F

  Filipe Martins Tendeiro

  ist1108325

Supervisors of this institution:

• 

  A

  Alexandre José Malheiro Bernardino

  ist13761

• 

  N

  Nuno Alexandre Antunes Martins Pessanha Santos

  ist157680

electrical-engineering-electronic-engineering-information-engineering - Electrical engineering, electronic engineering, information engineering

eng - English

Instituto Superior Técnico