LITESENSE

With the rise of Artificial Intelligence and the IoT, edge computing is becoming increasingly important in many sectors. However, current solutions are limited by their energy consumption and bandwidth. LiteSense meets this need by offering lightweight, high-performance alternative software that can be deployed on constrained hardware.

A unique algorithmic approach that challenges traditional pyramid-based one-shot vision AI architectures and enables their use with minimal resources (memory, energy), without compromising on accuracy and robustness. A flexible, hardware-agnostic innovation that can be integrated into multiple devices (autonomous cameras, robotics, IoT modules, smartphones).

Vision AIs that consume three times less computing power than state-of-the-art solutions. For end users:

• Increased energy efficiency (compatible with batteries, solar power, and embedded systems)
• Reduced infrastructure costs (fewer servers, less bandwidth)
• Immediate response time (local analysis, no network latency)
• Enhanced privacy (data processed at source, transmission limited to events)
• Operational resilience (works offline, ideal for isolated or critical sites)

Various detection demonstrators (people, gauges) have been implemented on an evaluation board containing a low-memory, low-power microcontroller. The TRL is 5.

• Monitoring of industrial infrastructure
• Personal safety
• Isolated or critical sites
• Constrained and mobile robotics, drones
• Logistics