FireGard is MidGard’s AI-powered wildfire detection platform designed to address critical gaps in early wildfire detection and rapid response. It replaces traditional image comparison methods with a proprietary deep learning algorithm trained on over one million annotated images of smoke and fire. These images, sourced from more than 1,200 cameras across North America and Europe—including operational deployments in Marseille, Gironde, Finistère, and Sologne—ensure robust performance in diverse and challenging conditions.
Developed entirely in-house over three years of intensive R&D, FireGard is tailored for early-stage fire detection, achieving high sensitivity with a low false positive rate (typically 5–10%). In 2023, its capabilities were enhanced by integrating a temporal analysis layer that evaluates image sequences, further reducing false alerts while maintaining rapid detection.
Today, FireGard is deployed across over 3 million hectares in France, with more than 300 cameras installed. MidGard is trusted by emergency services in more than 20 Fire and rescue services, and is strengthened by MidGard’s strategic partnership with Entente Valabre since 2021. In Gironde alone, 132 cameras have been operational since 2023. Beyond detection, FireGard includes a full-featured alert management platform: real-time image analysis, automated geolocation of ignition points (via triangulation and telemetry), intuitive visual interfaces, remote camera control for alert verification, and optional integration with NexSIS.
FireGard is a proven operational solution (TRL 9) that has been tested and qualified in real-life missions. It exemplifies innovation, bringing a technology known and used in the world for less than 3 years, yet uniquely adapted and optimized for the specific needs of fire and rescue services.
MidGard was founded with a clear ambition: to strengthen the resilience of territories facing growing environmental risks, particularly wildfires. In a context of accelerating climate change and increasingly intense fire seasons, our goal is to provide public safety services with the most advanced and reliable tools for early detection, rapid response, and coordinated action.
With FireGard, our flagship AI-based wildfire detection platform, we aim to redefine how forest fires are detected and managed. Our ambition is to accelerate detection speed, support firefighters in monitoring evolving situations, and minimize false alerts to maintain operational clarity. FireGard is not just a sensor network—it is a comprehensive decision-support tool that helps first responders act faster, smarter, and with greater confidence.
Our technology is the result of over three years of dedicated R&D, with a deep learning algorithm trained on a uniquely large and specialized dataset. By continuously enriching our dataset and refining our model, we ensure that FireGard remains the most precise and field-tested system available.
Looking ahead, our ambition is to expand FireGard across Europe and beyond, and to adapt it to a broader range of natural risks such as floods, landslides, and storms. By combining AI, video analysis, and sensor data, we aim to detect early warning signs of diverse hazards before they escalate.
This multi-risk approach is central to our strategy: building a unified, modular platform for real-time environmental threat monitoring. We’re also committed to further integration with national emergency platforms like NexSIS, and to developing lightweight, energy-efficient models for deployment in remote or low-connectivity areas.
MidGard envisions a future where technology enables proactive, data-informed civil protection, and FireGard is at the heart of this transformation.
FireGard is the result of over three years of targeted R&D and marks an important innovation in wildfire detection. Unlike traditional systems based on motion detection or pixel comparison, FireGard uses a proprietary deep learning algorithm trained on more than one million annotated images from varied natural environments. This training enables the system to detect the earliest signs of smoke with high accuracy—even in visually complex conditions like haze, low light, or cloud cover—while maintaining a low false positive rate of 5–10%.
The model’s performance has been validated through large-scale deployments across France. Feedback from field operations is continuously fed into our development cycle to improve the system’s precision and adaptability. In 2023, FireGard incorporated temporal consistency—analyzing sequences of images rather than single frames—which significantly reduced false alerts without compromising detection speed.
FireGard’s innovation goes beyond fire detection. Its modular AI framework is designed to evolve into a multi-risk monitoring platform, capable of addressing other environmental threats such as floods, landslides, or storms. By combining video analysis, sensor data, and AI-based pattern recognition, FireGard can support real-time detection of a wide range of natural hazards.
With features like remote camera control, automatic geolocation, and a web-based interface accessible from any device, FireGard is fully operational and can be easily integrated into national systems like NexSIS. It exemplifies how robust, field-tested AI can enhance civil protection and help communities anticipate and respond to environmental risks with greater agility.
FireGard responds to a key operational need: detecting wildfires early and supporting fire services throughout the incident. While traditional detection methods can be slow, imprecise, or labor-intensive, FireGard detects the first visible signs of smoke—often before flames appear—in under three minutes. Its deep learning model, trained on over one million annotated images, reliably identifies smoke in varied conditions (haze, clouds, low light), with a low false positive rate (5–10%).
But FireGard goes beyond detection. Once an alert is triggered, operators can take manual control of the camera that spotted the anomaly. They can zoom in, pan, or tilt the view in real time to confirm or rule out a fire. This doubt relief process avoids unnecessary deployments and allows teams to prioritize resources. During a confirmed fire, cameras can also be used to monitor fire spread in real time, helping commanders adjust strategy as conditions evolve.
FireGard includes a web-based platform designed specifically for fire and rescue services. It’s ergonomic and accessible from any connected device—PC, tablet, or mobile—and can be integrated seamlessly into emergency call centers. The interface provides a clear map view with live alerts, showing the location of all cameras and ongoing detections. Fire location is estimated using triangulation and telemetry data, with coordinates available for export or direct integration with other tools like NexSIS.
Fire and rescue services use the system daily to detect fires early, verify incidents quickly, and monitor situations as they unfold—improving both response time and coordination.
The FireGard project team brings together key experts: Anne-Sophie Cadre (President) oversees the project and deployments; Andrei Belokogne (CTO) leads technical execution; Michel Moukari (Head of R&D) drives algorithm innovation; Pierre Jardin (Infrastructure Lead) validates technical environments; and Damien Grandi (Web Lead) ensures the platform meets operational needs. This multidisciplinary team ensures seamless coordination and performance.
The FireGard project team brings together key experts: Anne-Sophie Cadre (President) oversees the project and deployments; Andrei Belokogne (CTO) leads technical execution; Michel Moukari (Head of R&D) drives algorithm innovation; Pierre Jardin (Infrastructure Lead) validates technical environments; and Damien Grandi (Web Lead) ensures the platform meets operational needs. This multidisciplinary team ensures seamless coordination and performance.