Home > News content

Ruixin micro AI chip plus Baidu PaddlePaddle to accelerate AI application landing

via:cnBeta.COM     time:2020/5/13 10:25:35     readed:91

In May 13th, Ruixin micro Rockchip officially announced that its AI chip RK1808 and RK1806 adaptable to Baidu PaddlePaddle (PaddlePaddle) open source deep learning platform, fully compatible with the lightweight propeller lightweight inference engine Paddle Lite.This time, Ruixin micro cooperated with Baidu to enable more application scenarios for AI industry and accelerate the landing process of AI products.

1.jpg

Baidu PaddlePaddle and Ruixin micro compatibility certificate

In the age of AI, deep learning framework is similar to operating system, which plays a role of connecting the preceding and the following, connecting chips and applications. AI chip with powerful computing power will gain more popularity.

NPUTime comes hard and softcombinationperformanceoptimization

RSIC micro AI chips rk1808 and rk1806 have built-in independent NPU neural computing unit, with int8 computing power as high as 3.0tops; with 22nm fd-soi process, the power consumption under the same performance is about 30% lower than the mainstream 28nm process products, with excellent performance in computing power, performance, power consumption and other indicators. The test results show that mobilenet V1 in paddy Lite takes only 6.5 MS to run, and the frame rate is as high as 153.8 FPS.

2.png3.png

Based on Baidu's years of in-depth learning technology research and industrial application, the open-source platform for in-depth learning of the propeller industry integrates core training and prediction framework, basic model library, end-to-end development kit, tool components and service platform. It was officially opened in 2016, and is an influential industry-level in-depth platform with comprehensive open source, leading technology and complete functions in China Learning platform. Paddy Lite is a set of lightweight reasoning engine with perfect function, strong usability and excellent performance, which supports a variety of hardware and platforms, and has important characteristics such as lightweight deployment and high-performance implementation.

4.png

Raytheon microRK18xxSerial chip adapts to paddy Lite

5.png

Raytheon microRK18xxSerial chips inMobileNETV1Top vs. mainstreamCPUExcellent performance

By adapting to the open-source deep learning platform of the flying oar, Ruixin microchip will be able to better meet the business needs of domestic users and provide strong computing power for the end-side AI; the integration of the two will give full play to the advantages of the combination of software and hardware, speed up the development and deployment, and promote the landing of more AI applications.

Detailed explanation

Refer to the file of paddy Lite for the detailed operation method of Ruixin micro AI chip on the flying oar, including the supported chips, equipment list, paddy model and operators, demonstration of reference examples, etc.

(search path: baidu searches "paddy Lite document", selects release-v2.6.0 in the lower left corner, and deploys the case section "paddy Lite uses rk NPU to predict deployment")

6.jpg7.jpg

Test equipment(RK1808 EVB)

8.png

In addition, in addition to RK1808 and RK1806 chip solutions, Ruixin micro NPU's AI chip will also be upgraded to match Baidu PaddlePaddle, further deepen bilateral cooperation, and work together to help build our own controllable AI ecosystem.

China IT News APP

Download China IT News APP

Please rate this news

The average score will be displayed after you score.

Post comment

Do not see clearly? Click for a new code.

User comments