JIMU Intelligent, a leading intelligent driving technology provider, launched its latest vision-based ADAS solution JM600 V3.0 that based on Xilinx’s high-performance FPGA platform and JIMU’s self-developed outstanding deep learning technology, delivering the best balance between performance and costs. JM600 V3.0 is also said to be the most cost-competitive ADAS solution based on FPGA, about 30% lower in total cost than competitors.
Targeting at the pre-installed market, JM600 V3.0 is predicted to go into production in 2018 Q3.
Rich Safety Features, Leading Detection Accuracy
As for the system features, JM600 V3.0 driving assistance system is capable of detecting vehicles, lanes, pedestrians, cyclists, and speed limit signs ahead, and will issue warnings in advance after detecting collision dangers and risky driving behaviors like over-speeding.
With its self-developed leading deep learning algorithm, JM600 V3.0 is capable of detection of vehicles 120m-150m ahead, and the distance detection error is within 3% when compared with Lidar. Lane markings including solid lines, dashed lines, double lines, rumble markings, road edge, roadside fence can be accurately detected as well.
Besides, JM600 V3.0 could be integrated with radar to bring Level 2 intelligent driving functions like autonomous emergency breaking (AEB), lane keep assist (LKA), etc.
Powerful FPGA Platform, ISP Integrated
JM600 V3.0 solution runs on the powerful FPGA platform from Xilinx. Compared to ASIC and GPU, FPGA outruns competitors in its programmability and system flexibility. It could handle the increasing computation requirement of AI algorithm with its powerful parallel processing capability, and also address the different requirements on interface customization.
Additionally, JM600 V3.0 comes with ISP integrated, which supports HDR to deliver better image quality, and more importantly, reduces the solution’s total cost.
To facilitate mass production, JIMU Intelligent has done a lot of work to control the solution cost, including improving the system integration, selecting more cost-efficient FPGA platform, and JM600 V3.0 ends up being the most cost-competitive FPGA-based ADAS solution released, about 30% lower in cost than its rivals.
Excellent Deep Learning Technology
As mentioned before, JM600 V3.0 is equipped with the latest deep learning algorithm from JIMU Intelligent. To make deep learning suitable for FPGA platform, the deep learning network should be compressed, and its power consumption should be reduced.
Through encoding pattern optimization, redundant network cutting, and algorithm flow adjustment, JIMU Intelligent successfully reduced the deep learning model size and computation by 90% while the accuracy loss is less than 1%.
JIMU Intelligent, Ranking No.1 in Large Coach Customer Share in China
By virtue of its overall advantage in leading technology and technical support system, JIMU Intelligent has achieved cooperation with over 50% of TOP15 coach OEMs in China, including King Long (Suzhou), Golden Dragon (Xiamen), Weichai Power, Sunlong(Shanghai), Youngman Automotive, Yinlong New Energy, Wanxiang (Shanghai), etc, ranking No.1 in terms of large coach customer share in China. At present, driving assistance systems from JIMU Intelligent are fitted to vehicles from these manufacturers as standard.
About JIMU Intelligent
JIMU Intelligent, founded in 2011, is a leading AI company that dedicated to vision-based intelligent driving technology. Based on its self-developed computer vision technology, JIMU is always in pursuit of the development and popularization of intelligent driving. At present, JIMU has established cooperation with many car manufacturers to bring about future smart driving and transportation, and forged partnership with insurance, fleets, and intelligent transportation firms to release cost-efficient ADAS-based industry solutions.