FusionSense — adaptive multimodal fusion at the sensor edge
Most edge systems either fuse sensor data on the cloud or filter each modality separately, wasting bandwidth and missing cross-modal patterns.
FusionSense trains lightweight classifiers at the sensor itself to decide what to transmit in real time, aware of dependencies between cameras, LiDAR, and depth streams.
Cuts energy and latency for autonomous systems and smart-industry deployments where tight compute budgets and reliability demands rule out naive fusion.