Early wildfire detection is critical for effective suppression efforts, necessitating rapid alerts and precise localization. While computer vision techniques offer reliable fire detection, they often lack contextual understanding. This paper addresses this limitation by utilizing Vision Language Models (VLMs) to generate structured scene descriptions from Unmanned Aerial Vehicle (UAV) imagery. UAV-based remote sensing provides diverse perspectives for potential wildfires, and state-of-the-art VLMs enable rapid and detailed scene characterization. We evaluated both cloud-based (OpenAI, Google DeepMind) and open-weight, locally deployed VLMs on a novel evaluation dataset specifically curated for understanding forest fire scenes. Our results demonstrate that relatively compact, fine-tuned VLMs can provide rich contextual information, including forest type, fire state, and fire type. Specifically, our best-performing model, ForestFireVLM-7B (fine-tuned from Qwen2-5-VL-7B), achieved a 76.6% average accuracy across all categories, surpassing the strongest closed-weight baseline (Gemini 2.0 Pro at 65.5%). Furthermore, zero-shot evaluation on the publicly available FIgLib dataset demonstrated state-of-the-art smoke detection accuracy using VLMs. Our findings highlight the potential of fine-tuned, open-weight VLMs for enhanced wildfire situational awareness via detailed scene interpretation.
Early wildfire detection is critical for effective suppression efforts, necessitating rapid alerts and precise localization. While computer vision techniques offer reliable fire detection, they often lack contextual understanding. This paper addresses this limitation by utilising Vision Language Models (VLMs) to generate structured scene descriptions from Unmanned Aerial Vehicle (UAV) imagery. UAV-based remote sensing provides diverse perspectives of potential wildfires, and state-of-the-art VLMs enable rapid and detailed scene characterization. We evaluate both cloud-based (OpenAI, Google DeepMind) and open-weight, locally deployed VLMs on a novel evaluation dataset specifically curated for forest fire scene understanding. Our results demonstrate that relatively compact, fine-tuned VLMs can provide rich contextual information, including forest type, fire state, and fire type. Specifically, our best-performing model, ForestFireVLM-7B (fine-tuned from Qwen2-5-VL-7B), achieves a 76.6% average accuracy across all categories, surpassing the strongest closed-weight baseline (Gemini 2.0 Pro at 65.5%). Furthermore, zero-shot evaluation on the publicly available FIgLib dataset demonstrates state-of-the-art smoke detection accuracy using VLMs. Our findings highlight the potential of fine-tuned, open-weight VLMs for enhanced wildfire situational awareness via detailed scene interpretation.
Wildfires are an increasing global threat, exacerbated by climate change. Traditional detection methods, such as ground patrols and satellites, often face delays, limited coverage, or high costs. This paper presents Evolonic, an autonomous drone-based wildfire detection system developed by Fraunhofer IISB, Fraunhofer IIS and Friedrich- Alexander University Erlangen-Nuremberg. The system integrates a fixed-wing VTOL UAV, computer vision- based smoke detection, a web-based alert platform, and an automated base station for continuous operations. A comparative analysis with existing detection technologies (sensor networks, cameras, and satellites) highlights the advantages of UAV-based monitoring in detection speed, flexibility, and real-time verification. Initial results confirm its potential for rapid wildfire detection and improved emergency response. While challenges remain, such as regulatory constraints and nighttime detection, this research demonstrates UAVs' role in enhancing wildfire monitoring and response, offering a scalable and adaptable solution to mitigate wildfire risks.