VeriSilicon (688521.SH) today announced AcuityPercept, its AI-based automatic Image Signal Processing (ISP) tuning system, designed to intelligently optimize image processing parameters for enhanced object recognition. AcuityPercept improves the accuracy and efficiency of AI perception systems by dynamically optimizing ISP parameters through automated tuning processes. It is widely applicable to AI-powered vision applications across various industries, including autonomous driving, robotic vision, and AIoT.
VeriSilicon’s AcuityPercept leverages a global directive and local refinement algorithm, utilizing metadata and loss feedback from AI task models to automatically achieve optimal ISP tuning. By continuously refining ISP settings, it effectively enhances object detection accuracy and delivers optimal ISP-processed images for neural network processing. AcuityPercept delivers a comprehensive optimization solution with VeriSilicon’s ISP IP, enabling automated closed-loop optimization.
“AcuityPercept is a key ISP tuning system that bridges the ISP with the target AI perception engine to achieve optimal object recognition. As AI-driven perception becomes increasingly critical in autonomous vehicles, smart surveillance, and robotics, our AI-powered automatic ISP tuning system lays the foundation for more accurate, efficient, and scalable AI vision solutions,” said Weijin Dai, Chief Strategy Officer, Executive Vice President, and General Manager of the IP Division at VeriSilicon. “AcuityPercept has demonstrated excellent results with road data collected by our automotive ISP customers. It is a crucial technology that enables vision perception engines to clearly and reliably identify objects.”
About VeriSilicon
VeriSilicon is committed to providing customers with platform-based, all-around, one-stop custom silicon services and semiconductor IP licensing services leveraging its in-house semiconductor IP. For more information, please visit: www.verisilicon.com
View source version on businesswire.com: https://www.businesswire.com/news/home/20250325468881/en/
