MAVNet is a compact neural network designed to run semantic segmentation on micro aerial vehicles with limited onboard computing resources. Its results illustrate how model design, task-specific datasets and evaluation metrics shape the trade-off between speed and accuracy.
Articles Tagged: embedded vision
Embedded Vision at the Edge: Faster Device Development
Embedded vision is expanding beyond specialized applications into industrial monitoring, quality management, robotics and more. A modular development approach can help engineering teams connect sensors, process image data and prototype devices while reusing existing hardware and software.