[Shanghai, China, September 26, 2016] Computer vision is a key component of the cognitive computing that will enable future communications networks, robotics and much more, but it will take significant new investment in R&D to meet the challenges of taking machine learning to the next step, according to Professor Liu Jianzhuang of the Huawei Shannon Cognitive Computing Laboratory.
In a keynote speech at the 2016 Symposium on Research and Application in Computer Vision (RACV2016), held at ShanghaiTech University, Prof. Liu spoke on Cognitive Computing – Challenges in the Industrial Application of Computer Vision to give an overview of the latest research in computer vision. Liu noted that great progress has been made with computer vision technologies based on deep learning, but he emphasized there is still a long way to go before deployment as a mature industrial technology.
“We need to invest more heavily in research on unsupervised learning, small sample learning, and biological cognition to significantly improve the reliability of machine intelligence,” Liu said.
Cognitive computing uses powerful algorithms to enable high-volume data analytics for perceiving people, things, events, and the environment, and supporting fast decision-making. It will be the key to future communications networks, data centers, smartphones, robotics, Internet of Vehicles (IoV), and other industries. Computer vision is one of the key technologies for cognitive computing. The goal is to simulate human vision, so that computers can perceive and complete tasks more efficiently and precisely. Computer vision is already in use in certain domains, and can greatly boost the competitiveness of some products.
Liu, who is chief researcher in computer vision at the Shannon Cognitive Computing Laboratory, said, "We are witnessing rapid development in computer vision applied to Internet image recognition using deep learning. Deep learning is now the area where researchers are focused, and an increasing number of industries are now aware of its importance. However, the mechanisms of deep neural networks still defy explanation, and the existence of adversarial images (slightly distorted images which computers cannot read) is a challenge to their deployment in real world applications."
The difficulties in understanding and controlling deep neural networks means that computer vision cannot yet meet demands placed on it by industrial applications. A key future research direction will be machine learning with more stability than currently available technologies.
Huawei established the Shannon Cognitive Computing Laboratory in early 2016. The Laboratory is helping to develop an intelligent world by making cognitive computing technologies like computer vision more stable and application-ready, to speed their commercial use.