AAA weekly

2020-09-07

SenseTime Advances Development of Self-driving Technology

China’s SenseTime, which researches and develops image recognition and AI technologies, has been rapidly developing since its foundation in Hong Kong in October 2014. In September 2019, SenseTime’s CEO Xu Li announced that the company's corporate valuation exceeded 7.5 billion USD. According to media reports, sales in 2019 reached about 750 million USD. Apart from the automotive sector, the company's involved in various areas such as smartphone apps and games, city safety management, medical care and education. In the automotive sector, SenseTime’s technology mainly used in driver monitoring and driver assistance systems for smart cockpits.

SenseTime began supplying driver monitoring systems to Chinese automaker WM Motor’s EX5 model in 2018. In April of the following year, SenseTime and Visteon concluded an agreement to jointly develop smart cockpits. SenseTime’s driver identification and fatigue detection technologies will be installed in Visteon’s cockpit products. In June of the same year, Shi Jun, President of the Asia Pacific Business Group of SenseTime as well as the Vice President of Strategic Planning of the Group revealed that the company will supply AI-base driver monitoring systems to 20 major automakers from the US, Germany, Japan and China until 2021.

SenseTime began research in the autonomous driving field in 2016. As of July 2020, SenseTime was conducting autonomous vehicle testing in China’s Hangzhou in Zhejiang province and Shanghai, and Japan’s Joso in Ibaraki prefecture. In Japan, the company is jointly developing autonomous driving systems with Honda. Mass production of the system is planned to begin by 2025.

The company is actively establishing overseas bases to increase overseas sales. After setting up an office in Japan in 2016, and in Taiwan in 2020, SenseTime set foot in Singapore in 2018, in the UAE in 2019 and in Taiwan in 2020.

SenseTime company logo

SenseTime: Company Information

Company Overview

Company name: SenseTime
Established: October 2014
Location of headquarters: Hong Kong
Additional locations: China (Shanghai, Beijing, Shenzhen, Chengdu, Hangzhou), Japan (Kyoto, Tokyo), Singapore, Taiwan, UAE, etc.
Products: Starting with image recognition, company is present in the number of fields developing various systems using deep learning.

History

・In Jul. 2001, Tang Xiao’ou, professor at the Chinese University of Hong Kong, set up the Multimedia Laboratory (MMLab). ・In 2011, MMLab applied deep learning technology to computer vision. ・In Sep. 2014, MMLab became second after Google at the ImageNet object recognition contest. ・In Oct. 2014, SenseTime Group was established in Hong Kong. ・In 2015, SenseTime won first place of the ImageNet Large Scale Visual Recognition Challenge (ILSVRC) which is a video object recognition contest. ・In 2016, SenseTime took first place in three categories of ILSVRC, namely “Object Detection,” “Object Detection and Tracking” and “Scene Parsing.” ・In Jan. 2016, SenseTime set up its first overseas company in Japan, called SenseTime Japan. ・In Sep. 2016, SenseTime concluded a business tie-up with US tech company Nvidia to develop smart cities. ・In Dec. 2017, the company signed a five-year long-term joint development agreement with Honda to establish autonomous driving technology in urban areas equivalent to Level 4 and make full-scale entry into the advanced driver assistance system (ADAS) market. ・In Apr. 2018, SenseTime raised 600 million USD from Alibaba and other investors. ・In May 2018, SenseTime raised 620 million USD from Tiger Global and other investors. ・In Sep. 2018, SenseTime raised 1 billion USD from SoftBank. ・In Aug. 2019, China headquarters and global R&D headquarters were set up in Shanghai. ・In Sep. 2019, SenseTime’s corporate valuation exceeded 7.5 billion USD.

Technology Overview

・Face recognition technology: Face detection / tracking, feature point detection, face matching, face attributes, face grouping, spoofing prevention, face editing ・Smart surveillance security: Pedestrian / vehicle detection, pedestrian attributes, crowd analysis ・Image identification: clothing attribute analysis, object detection, scene identification, vehicle type identification ・Script recognition: Receipt, credit card, license plate recognition ・Image / video editing: Fog removal, high resolution, noise removal, brightness correction, HDR correction, defocus correction ・Deep learning framework: Building deep learning algorithms using high CPUs and GPUs.

Automobile Technology Overview

・Smart cockpit – Face identification: Confirms the identity the driver of a ride-sharing vehicle and launches the air conditioner and telematics settings to suit each driver’s preference using face recognition – Driver monitoring: Detects inattention, yawning, inclination of the head, snoozing, smoking, eating, drinking and telephone use among other activities, and warns the driver – Gesture identification: Recognizes in-vehicle device operation ・Driving support – Lane departure warning: Applied when driving at 60 km/h or more. – Forward collision warning: Applied when driving at 200 km/h or more. Calculates the distance to the vehicle ahead and warns the driver 2.7 seconds before a potential collision occurs. – Pedestrian detection warning: Applied when driving at 40 km/h or more. Calculates the distance to the pedestrian and warns the driver 2.7 seconds before a potential collision occurs. – Distance detection warning: Applied when driving at 30 km/h or more. Detects and displays the distance to the vehicle ahead. – Traffic sign recognition: Identifies 14 types of signs. – Traffic signal recognition Chinese road signs (Created using SenseTime’s PR materials and various media sources)
Please log in or subscribe to AAA weekly to view the rest of this report.
Free trial subscriptions are available!