Seoul, Korea – May 7th, 2018 -- Chips&Media Inc., a leading global video IP provider, announced the launch of its first Computer Vision IP, detecting objects with a capability to process 4K resolution at 60 frames per second input in real-time. The IP implements Convolution Neural Network (CNN) for object detection and is differentiated by introducing fully hardwired acceleration and dedicated architecture to significantly reducing memory accesses and memory bandwidth requirements. Thereby, it maximizes the performance and minimizes power consumption remarkably. These characteristics will position Chips&Media’s computer vision IP ideal for computing-intensive edge devices in automotive and surveillance applications, where high performance, power, and area efficiencies are key.
This newly introduced IP will debut at Embedded Vision Summit 2018 to be held between May 21st and 24th at Santa Clara, CA. The showcase will include real-time object detection and lens distortion correction (LDC) running on an FPGA board with a target performance of 1080p @ 30fps to reflect the FPGA computing limitation.
The source from Chips&Media commented, “This is a very important moment as we are putting in the last piece of the puzzle to transforming ourselves as a total video IP provider, which we have prepared for a long time.” And added, “This is the last piece in the puzzle, but also marks as our first step as total video IP provider with IP catalogs including video codec, image signal processing, computer photography and finally the computer vision.”