On the morning of September 6th, 360 and Didi reached a strategic cooperation around the company's safety guarantee and special car service. The DDT will be equipped with a 360 driving recorder, and 360 will also cooperate with third parties to upgrade its operating system for 360 OS for car. .
In fact, at the 2016 CCF-GAIR global artificial intelligence and robotics summit held in Shenzhen in August, Yan Shuicheng, chief scientist of the 360 ​​Artificial Intelligence Research Institute and well-known expert in computer vision and deep learning, accepted Lei Feng network (search for "Lei Feng Net" public attention In the interview with ), he talked about his insights on computer vision.
In particular, he mentioned two situations in the interview:
1. Some smart cameras provide a function - to see if suspicious people enter the surveillance area, this function is generally used for facial recognition through computer vision technology. However, if the device is sold at a large scale, a large number of users use this function at the same time. If the calculation is performed on the server side, the server will be greatly pressured.
2. Many smart devices have no network application environment and cannot interact with the server.
To solve these problems, there are two directions:
Reduce the accuracy of calculations
For example, recognizing human face and age at the mobile phone end, accuracy is inevitably difficult to compare with a professional identification system. But this loss of precision from 95% to 85% is something people can afford.
Improve the calculation model
Of course, the best situation is to be able to develop new and better computing models. However, under the existing model, it is also possible to implement algorithm simplification by improving the strategy. For example, Yan Shuicheng led the team to study some algorithm adjustments and added some strategies based on the original algorithm to judge those logics without calculation. Although it seems that the rules are more complicated, the overall amount of computation is reduced.
Yan Shuicheng believes that the optimization of these strategies is very meaningful because costs are often an important factor in achieving commercialization. At present, the most mainstream artificial intelligence chips use GPUs. However, the GPU has a large volume and high energy consumption. If the cost reduction is achieved, it will be easier to apply it to practice.
At present, the 360 ​​Artificial Intelligence Research Institute uses the camera of the driving recorder to collect data and applies computer vision technology to the camera. Due to the robustness of the deep learning algorithm, the data collected by different cameras can help improve the effect. The resulting deformation is also not sensitive.
Deep learning is applied to driving recorders to provide drivers with reasonable and safe driving recommendations by testing driver behavior data . On the other hand, accident data can also be analyzed to provide vehicle owners with different road conditions, vehicle models and driving conditions. Timely driving assistance.
At this stage they have implemented some functions on low-cost, low-power ARM platforms.
On September 6, 360 and Didi traveled to cooperate on the safety and security of the company and the quality of the car service. DDT will be the first to equip 360 vehicle recorders to ensure traffic safety. 360 launched its first smart traffic recorder last year and has so far sold more than 3 million units to become the industry champion. At present, the 360 ​​driving recorder has become the official recommended driving recorder brand. At the same time, the 360 ​​Driving Recorder will be the first to be equipped with the DDT. Follow-up will also be explored in cooperation with hardware and software customization.
At the conference, the 360 ​​Institute of Artificial Intelligence introduced and demonstrated 360's layout and achievements in the area of ​​smart travel from the professional perspectives of ADAS assistive technology and image recognition.
ADAS, Advanced Driver Assitance Systems referred to as ADAS, Chinese name: Advanced Driver Assistance System. ADAS uses sensors installed on the car (cameras, radars, lasers, ultrasonics, etc.) to sense the surrounding environment, gather data, identify, detect and track static and dynamic objects, and combine vehicle travel. Data, systematic calculations and analysis are performed to alert the driver in advance of the perceived danger. )
The onboard camera is an important part of the ADAS (advanced driver assistance system). In combination with sensors such as radar and GPS, part of the automatic driving functions can be implemented, such as lane departure warning, forward collision prevention, and blind spot detection. The application of deep learning to ADAS is a trend. It has significant advantages in object detection and pedestrian recognition with insignificant features, but the disadvantage is high cost and often requires running on the server.
We can see that by applying computer vision to driving recorders and using deep learning techniques in ADAS assistive technology and image recognition, it is possible to better assist drivers in judging driving advice and driving assistance.
Related Reading
360 reached cooperation with Didi, and for the first time exposed 360 artificial intelligence research team
AIR 025 | 360 Chief Scientist Yan Shuicheng: Six Insights on Machine Vision
film capacitor,Snubber Capacitor,Dc Link Capacitor,High Current Capacitor
XIAN STATE IMPORT & EXPORT CORP. , https://www.capacitorhv.com