Compal’s CDAT AI perception stack is based on deep learning algorithms for object detection, combined with object tracking and action prediction to obtain object information. Currently, the algorithms can recognize object categories such as pedestrians, motorcycles, cars, bikes, trucks, and buses. The system enables ADAS functions, including Pedestrian Automatic Emergency Braking (P-AEB), to avoid a collision. Compal’s CDAT perception stack supports ADAS features such as forward collision warning, Level 2 and 2.5 hands-free driving, Level 3 eyes-off hands-off driving, and Level 4 full-self-driving tasks.