Just recently, Baidu has chosen Inspur’s NF5568M4 server to give hardware support for the deep learning platform of its autonomous car, which is one of the key areas of driver technology.
In December last year, the Chinese company’s driverless car passed the road test beginning from the Baidu Tower in Zhongguancun High-tech Park in Beijing, driving onto the G7 Beijing-Xinjiang Expressways, through the Fifth Ring Road and reaching Olympic Forest Park, and back.
Included in the driverless technology embraced throughout the course are lane switching, decelerating, overtaking, going on and off ramps and turning around. The car successfully passed each of these testing criteria and reached a top speed of 100 kilometers per hour.
Image sensing and recognition is another key technology in driverless cars. This enables Baidu’s car to precisely sense objects and pedestrians, even efficiently following traffic lights and preventing accidents. Image recognition is an ultra large computing project, which needs hundreds of thousands—in some cases even billions—of learning samples to train a model. Hence, a GPU made up of thousands of smaller and more energy-saving cores has become the key force for the application of image recognition training.
Inspur is a leader in artificial intelligence application and has strategic collaborations with NVIDIA—the largest vision computing firm in the world—in the field of GPU heterogenous computing. Aiming to produce the high-performance computing application, Inspur has unveiled the NF5568M4 co-processing acceleration server, carrying four Nvidia Tesla K40 GPUs and two Intel E5-2600v3 processors. The highest single computing capacity of a single GPU server reaches 17 teraflops.
Currently, in the KITTI test for common vehicles, Baidu has reached a recognition accuracy of 90 percent—thanks to a huge contribution by Inspur’s GPU co-processing acceleration server NF5568M4, which ensures greater safety in the operation of autonomous cars.
Baidu intends to set up demo regions in 10 cities within China and put commercial-purposed autonomous cars to use. Huge scale projection is anticipated in five years, and within 15 years, it is projected that 80 percent of newly produced cars will be equipped with the driverless function.
Insur comes with 100,000-core-and-above CPU+GPU/MIC/FPGA large-scale paralleling algorithms, software optimization and program development abilities.
Moreover, through research and development, Inspur’s open-source Caffe-MPI and ClusterEngine high-performance computing management platform are centered on deep learning and artificial intelligence to establish a professional heterogenous acceleration platform. This platform enables the debut of deep learning applications such as autonomous cars and other groundbreaking applications.
Hamid Moaref has always been fascinated by cars and the automotive industry. His family has a longstanding association with the industry and has been in the tire business for the past 35 years. Raised in Dubai, Hamid attended Capilano University in Vancouver where he graduated with a BBA in marketing before attending an intensive course in magazine publishing in 2005. He has been the publisher and chief editor of Tires & Parts magazine for the past ten years.
Your email address will not be published. Required fields are marked *
© 2017 Morjan Media LLC. All Rights Reserved.