Open Rescue: a technological and human adventure

Open Rescue was created to help locate people in distress more quickly through the automated analysis of drone-captured images. But contrary to what one might think, this technology wasn't developed in isolation, nor solely behind a screen. It's the result of fieldwork—collective and, above all, human.

A bird’s-eye view to see further: aerial image capture

It all begins in the air. Our drones, equipped with cameras and sensors, fly over varied terrains—dense forests, mountainous areas, rugged trails—to record a wide range of situations representative of real-life search and rescue missions.

But above all, these image captures were made possible thanks to the valuable collaboration of the Fédération Française de Randonnée 74, whose members played the role of victims to simulate realistic scenarios. A heartfelt thank-you as well to SDIS 74, who didn’t just attend the test sessions, but actively supported us throughout the entire project. Their expertise and commitment were crucial at every stage, from designing scenarios to carrying them out in the field.

A huge thanks to them for their on-the-ground involvement—without them, the AI wouldn’t be able to “see” much at all!

The key role of humans: image annotation

The collected images aren't magical—Open Rescue needs to learn, and that means every single figure must be spotted, tagged, and annotated by hand. This step, known as data labeling, is essential for training the AI model to detect a person on the ground, even when partially hidden by vegetation.

Machine learning: when Open Rescue takes over

Once the data is prepared, Open Rescue gets to work: learning to recognize human shapes in brand-new images. It compares, analyzes, adjusts, and eventually learns to identify signs of human presence—on its own—even in complex, cluttered environments.

Trial by fire: real-life testing

And because AI means nothing without real-world conditions, Open Rescue was put to the test in full-scale field exercises. This crucial phase of the project was carried out with the invaluable support of SDIS 74, who provided their expertise, supervision, and resources to ensure that testing conditions were as close as possible to real-life rescue operations.

Thanks to their contribution, we were able to:

• Validate the AI’s detection capabilities

• Assess the system’s responsiveness

• Adjust operational needs based on field feedback

AI built to assist—not replace—rescue teams

Open Rescue is not designed to replace rescue teams. Its purpose is to save them valuable time by sorting and prioritizing areas to search.

A drone flies overhead, Open Rescue analyzes the footage in real-time, and responders can focus on the zones with the highest probability of human presence.

The development of Open Rescue is, above all, the story of a close collaboration between technology, rescue professionals, and people on the ground.