Baidu taps Intel VPUs to power AI retail camera

The new Xeye camera, which uses Intel Movidius VPUs, is one of multiple products from the Chinese search engine giant that leverages Intel's AI technology.
Written by Stephanie Condon, Senior Writer

Intel on Tuesday announced that Baidu is using Intel Movidius VPUs to power a new AI camera for retailers, called the Xeye.

The Chinese search engine giant is also using Intel technology to power various other artificial intelligence products and services, the companies announced at Baidu Create, Baidu's AI developer conference.

The Xeye combines Intel's Movidius Myriad 2 vision processing units with Baidu's machine learning algorithms to analyze objects and gestures, as well as to detect people, to provide personalized shopping experiences in retail settings. AI has become a key component of digital initiatives within the retail sector.

Meanwhile, Baidu is tapping Intel's FPGAs (field-programmable gate arrays) to offer workload acceleration as a service on the Baidu Cloud. Baidu has also optimized PaddlePaddle, its deep learning framework, for Intel Xeon Scalable processors.

Additionally, the two companies announced a new partnership focused on commercializing safety systems for autonomous vehicles. Baidu plans to integrate and commercially deploy the Mobileye Responsibility Sensitive Safety (RSS) safety model in its open source Project Apollo as well as commercial Apollo Pilot programs.

This makes Baidu the first autonomous driving solution provider to publicly adopt RSS -- a safety model that adds a deterministic decision-making layer on top of AI-based safety systems. Deterministic systems ignore uncertainty; in terms of self-driving vehicles, that means the model formalizes certain safe driving concepts like following cars from a safe distance. Baidu and Mobileye (an Intel subsidiary) will jointly work on updating the RSS model for the unique driving styles and road situations in China.

Additionally, Baidu is adopting Mobileye's Surround Computer Vision Kit, which includes 12 cameras positioned around a vehicle plus Mobileye's computer vision hardware and software.

Editorial standards