The eyes of Intelligent Cities

In the future, with the help of 5g the intelligent city will focus on the AIoT for intelligent infrastructure, like using AI to recognize license plates for parking fees, AI policing with drones, and etc. All these applications need the base of computer vision.

In 27 March 2020, Huawei webcasted the Huawei Developer Conference 2020 (Cloud) in Shenzhen. During the conference, Huawei unveiled the Huawei Computer Vision Plan which with main directions of how to efficiently mine useful information from massive amounts of data, design efficient visual models for identifying everything, and express and store knowledge to move towards universal intelligence, incorporate 6 sub plans,

• Data Iceberg: Use a small amount of labeled data to pry up massive unlabeled data to support model training in small sample scenarios
• Data Magic Cube: use multiple modals to assist each other and enhance the model’s learning ability in actual scenarios
• Model High Touching: build a large cloud-side model and refresh the performance ceiling of various visual tasks
• Mode Slimming: Efficient computing models on the device side will help various chips complete complex inference.
• Generic Vision: Vision pre-training tasks will build generalized vision models.
• V-R Integration: Direct computer vision will create real artificial intelligence

From human intelligence to artificial intelligence, it depends on how machines perceive the world. Computer vision is not just a way to improve smart products, for cities, computer vision is the city’s eye for internal activity and a means of responding to changes brought about by new products, such as autonomous driving.