Neurosoft, in cooperation with AI-focused Israeli company Hailo AI, has performed an implementation of its own specially developed neural networks for vehicle and pedestrian detection on the Hailo-8™ neural coprocessor. This made it possible to significantly increase the performance of key NeuroCar functions in on-edge applications.
The Hailo-8™ neural coprocessor enables highly efficient image processing at up to 800 frames per second. The solution enables simultaneous processing, in real time, of a video stream from 50 cameras for vehicle and pedestrian detection and tracking, using a single Hailo unit with a power requirement of just 4 W.
The new processor has found applications in NeuroCar systems based on vehicle and pedestrian tracking, such as. NeuroCar Park Spots (parking space occupancy determination), NeuroCar RedLight (red light crossing detection), NeuroCar Incident Detection (detection of dangerous incidents on the road) and NeuroCar Pedestrian Crossing (pedestrian crossing safety monitoring).
The second major field of application for the Hailo-8™ processor is to use it for batch processing on servers in the Neurosoft cloud. This allows to significantly increase the performance of NeuroCar systems in off-edge applications, such as for analyzing video recordings of traffic measurements.