Making Sense of Sensetime
By Dr Li Xu, Co-founder & CEO of SenseTime | Sensetime was set up in 2014, and is China’s largest artificial intelligence (AI) unicorn focused on computer vision and deep-learning technologies. Prof Xiao’ou Tang and a team of 10 scientists and engineers including myself are the Co-founders. Headquartered in Hong Kong, SenseTime has offices in Beijing, Shenzhen, Shanghai, Hangzhou, Chengdu, Kyoto and Tokyo.
Our target definitely is not to create a small company to be acquired, but rather a ‘platform company’ for global users that dominates with original core technology, like Google and Facebook.
China’s Largest AI Unicorn
SenseTime is focused primarily on computer vision and deep learning, and has independently developed a deep learning platform and a deep learning supercomputing center. With proprietary technologies serving as its foundational driver, SenseTime is China’s largest AI unicorn (valued at more than US$3 billion) and has established an R&D center, integrated with various industries, and forged partnerships across the board to create an AI ecosystem.
Wide Range of Technologies
Over the past three years of development, SenseTime has launched a wide range of AI technologies including facial recognition, image recognition, text recognition, video image analysis, image and video editing, remote sensing, autonomous driving, medical image recognition, and AI computer chips. In doing so, we have become China’s largest AI algorithms provider.
Working With HKSTP
We established SenseTime at Hong Kong Science & Technology Parks (HKSTP) in Shatin and our headquarters is still there. HKSTP helped to arrange and facilitate networking, marketing and technology demonstration opportunities for SenseTime. For example, SenseTime was engaged in the visits of key government officials at HK Science Park, including the Chief Executive of Hong Kong, Financial Secretary, and senior government officials from the mainland.
HKSTP also facilitated business sharing sessions between SenseTime and major authorities such as HKMA and HKEx. Through the “Technology from Science Park” programme, HKSTP has arranged one-on-one business matching for SenseTime and large enterprises.
Deep Learning Platform
SenseTime’s independent and proprietary deep learning platform Parrots is optimized to support wide-scale networks, extra-large data study, and complex applications. To speed up the time it takes to develop deep learning algorithmic models, SenseTime has built a deep learning supercomputing center that significantly reduces R&D costs for a wide range of AI technologies. It effectively serves as SenseTime’s hardware infrastructure for algorithms R&D.
In the last few years, SenseTime’s computer vision technologies have quickly landed in the market. They include facial recognition, image recognition, text recognition, video image analysis, image and video editing, autonomous driving, remote sensing, and medical image recognition.
Besides its technological strengths, SenseTime has achieved commercial successes. It has powered many industries, including security, smart phone, mobile internet, automobile, financial services, retail, education, and health care.
Customers & Corporate Partners
SenseTime has served more than 400 customers and partners in China and abroad, among them world-renowned companies such as Honda, Qualcomm, NVIDIA, China Mobile, UnionPay, HNA, Suning, Huawei, Xiaomi, OPPO, vivo, Weibo and iFLYTEK. As such, SenseTime’s solutions cover security, finance, automobile, smart retail, smartphone, mobile Internet, robotics, and other industries.
SenseTime has crafted the unique and successful “1 (basic research) + 1 (products and solutions) + X (industry)” model that uses SenseTime’s engine to get results for the industries. SenseTime currently leads the market in almost all of its product categories, as well as industry commercial revenue. It also leads in the many vertical industries it serves.
Contributing To Society
Here are two impressive cases that prove our solutions contribute value to society:
The high-security prison in Inner Mongolia leverages the national first dynamic face recognition people control system developed by SenseTime, which effectively assists the security control in the prison, increases management efficiency and enhances staff productivity.
With the use of high-definition videos, the system can recognize faces in real scenarios of day and night attendance, surveillance in major zones, staff patrol in prison, staff attendance, etc. The system is used mainly for authentication and intelligent security management to highly improve safety, precautions and the crisis handling ability of the prison.
One Public Security Bureau in China leverages SenseFace, a face detection platform developed by SenseTime that effectively assists police officers in detecting and finding lost elderly people within hours after family members report the case to the bureau. SenseTime contributes to the human value of family reunion. It greatly assists the bureau to save working time and solve problems in an efficient way.
Face detection systems can detect faces within milliseconds, reading even low-quality, complex pictures and surveillance videos of large crowds. They are capable of decoding facial profiles, partially obscured and blurry images, and moving facial expressions, as well as facial images taken in other environments.
Safer Self-Driving Cars
SenseTime recently announced a five-year R&D partnership with Honda involving self-driving cars. Safety is the number one priority when it comes to driving, and it also constitutes the core of our autonomous driving solution.
By combining SenseTime’s strengths in computer vision technologies with Honda’s superior vehicle control technologies, we will together enable a safe and pleasant autonomous driving experience anywhere, including urban areas. Moreover, the fact that SenseTime provides core technology to a global enterprise like Honda marks a milestone.
Lots Of Investor Interest
The country’s top venture capital and private equity firms have brought robust support. Following the latest financing round (US$410 million Series B and US$227 million from Alibaba), SenseTime has allocated more resources to developing autonomous driving, intelligent medical treatment, and deep learning computer chips. It will also strengthen its technology platform and attract more talent to open up greater applications scenarios as well as a SenseTime-driven AI commercial ecosystem.
What The Future Holds
We plan to do an IPO, but don’t have concrete timeline at this moment. Also, we are working on a US R&D center and will share more details when the time is right.
We have to attract more talents to open up greater applications scenarios, as well as a SenseTime-driven AI commercial ecosystem. Having a deep understanding of the vertical industries we serve, and keeping up as the industries evolve can also be challenging.
About 500 Patents, So Far…
SenseTime is one of the few Chinese companies that owns proprietary deep learning platforms, which gives us an edge over competition. We’ve filed about 500 patents, which is significant for a company that is just three years old. For any startups to succeed, we believe it’s important to own IPs and have the ability to develop proprietary technologies.
About The Author
Dr Li Xu is Co-founder and CEO of SenseTime, one of the unicorn startups in China. Under his leadership, Dr Xu has led SenseTime to achieve a number of breakthroughs in artificial intelligence R&D. At ImageNet 2015, SenseTime was the first Chinese company to win the top award. At ImageNet 2016, his team won three individual championships.
Prior to joining SenseTime, Dr Xu worked at Lenovo Corporate Research & Development, and was visiting researcher at Motorola China R&D Institute, Omron Research Institute, and Microsoft Research.
Dr Xu obtained bachelor’s and master’s degrees in computer science from the Shanghai Jiao Tong University. He holds a doctor’s degree in computer science from the Chinese University of Hong Kong. With doctoral research focusing on computer vision and computer imaging, Dr Xu has published more than 50 papers at top international conferences and in journals in the field of computer vision.