Every year people become lost or injured far from the trappings of civilization. Activating a Search and Rescue (SAR) operation can prove challenging due to harsh terrain and weather conditions, and another factor we now fortunately take for granted: reliable communications. The good news is that a team of SAR experts and key technology partners have assembled a solution that could prove a breakthrough in effecting a rescue, when hours or even minutes make a difference.
Intel 5G Innovation Center/DCAI Labs Stockholm showcased a live AI/video analytics solution for the attendees at the Search and Rescue event in Lom, Norway on Sept 22nd. This event followed the SAR mandate of finding people in the terrain and directing rescue personnel to the person in need of being rescued. But rather than using the traditional method of helicopter pilot and radio transmissions, this solution employs an RGB and IR video feed from drones within a private 5G network to boost effectiveness and extend the capabilities of the rescue team. Intel has been working on this project together with Telenor (5G-SA) and Norsk Luftambulanse (Drones) for about a year.
This event was the first milestone for the project, and the team received great feedback from the attendees coming from multiple SAR services. So far, staff have arranged rescue sceneries at different locations in the mountains to collect video feeds, data, pictures, interviews, and input from the people working in the field and videos for the next step in the project. This data will help to train AI models to identify features more accurately in the landscape, and to correlate audio communications with key SAR activities. These materials will also be used to create a new virtual demo in the 5GIC to give potential customers a more realistic exposure to the full solution. Future milestones will be focused on adding new features and capacities.
The architecture and software used in this solution include Intel third generation scalable Xeon® processor utilizing the DL boost VNNI instruction set to accelerate inference. In addition, the OpenVINO™ framework and custom code was configured to handle multiple video streaming formats. Intel NUCs were used to interface with third party API´s. Other development activities include using Intel annotation software and training several deep learning computer vision models. An MLOps pipeline from edge to cloud was deployed by continuously updating models used for detection and classification, demonstrating viability for real world applications.
So, it appears that when danger calls, technology really can come to the rescue. Visit the Intel 5G Innovation Center to learn more.
I am fascinated by the potential for AI and ML to transform business and society, and occasionally say interesting things about it. My educational background includes a CS degree, AI/ML post-grad work, and AWS certifications. I have served in AI marketing roles at IBM, Lenovo, and now Intel.