MACHINE VISION
We’ve seen amazing progress over the past decades with the emergence of sensors and image processors which can mimic the eye’s ability to ‘see.’
These technologies continue to improve in sensitivity and precision, enabling today’s cameras to capture images at great speeds, increasingly with built-in depth perception. They can analyze raw sensor data in real time, and generate digital representations that enable decision making.
Machine vision today is enabling a wide range of new use cases for applications including smart home monitoring, automotive safety, drones, retail analytics, and much more.
OST’s GP8300 SoC represents a new architecture that achieves deep integration of eAI, edge computing, and communications on a single chip. It provides the low-power, high-performance, heterogeneous processing power to help OEMs create the next generation of products that take advantage of machine vision.
Solution Benefits
-
Reduces chip cost, area, and power consumption for image recognition and object detection
-
Includes four 2GHz interconnected with a cache coherent memory supporting Heterogeneous Systems Architecture (HSA) processing for a common programming framework
-
Integrates four of GPT’s new 2GHz Variable Length Vector DSP (VLVm1) cores for signal processing applications
-
Integrates two 1GHz ‘Song’ AI accelerators from GPT for efficient Convolutional Neural Network (CNN) inference
-
Integrated functionality includes worldwide GPS
-
Incorporates multiple interfaces such as eSIM
-
Supports customization and software upgrades for NB-IoT standards updates
-
Supported by CaffeNet-based training and tools for automatic fixed-point conversion and compression for inference