AI face detection algorithms can extract features of faces and clothing and classify them into useful categories such as gender, age and clothing color. Searching this rich attribute information can help us easily find the target person, such as searching for pictures by face, controlling face, etc.
How to build a dynamic deployment and control system for facial recognition of key parts?
Through the TSINGSEE Qingxi Video AI intelligent analysis system, an automatic early warning system with dynamic face recognition is constructed to accurately capture the faces of people in the area monitored by the face recognition surveillance cameras connected to the system, and obtain clear faces. Images, the system automatically completes the archiving of portrait information, and can also compare it with face images in the key personnel database, thereby playing the role of capture, identification, early warning and prevention.
Demarcate the specified number of personnel in a specific area. Once the system detects that the number of personnel exceeds the fixed value, the TSINGSEE video AI intelligent analysis system can immediately issue an alarm and notify the backend management personnel. It can also be linked to voice control to connect to the scene in real time for voice prompts and alarms. .
The current face detection algorithm can achieve high-precision face detection, and is also effective for detection in low-light and backlight environments or for detecting faces obscured by masks.
It should be noted that in face detection scenarios, the installation position of the camera is also particularly important.
- The camera should be installed as close as possible to the horizontal position where the face enters the detection area.
- The camera should be installed no more than 15 feet from the facial recognition target area.
- The camera's downward angle should be 15 degrees or less.
- Face detection cameras should be installed 6 to 10 feet from the ground in the face detection area.
- The left or right viewing angle should be less than 30 degrees, and the straighter the angle, the better.
- When the camera is installed at the entrance of a building, the detection effect is better, because in such a scene, only one or a few people can enter the detection area at the same time.
In order to help developers find the best face dataset to meet their needs, in this issue we have also collected some commonly used and high-quality public datasets focusing on faces.
1）Digi-Face 1M
Released in 2022. Digi-Face 1M is the largest synthetic face recognition dataset, with 1.2 million face images.
2）VGG Face2
Released in 2018, the dataset contains 3.31 million images and is rich in pose, age, lighting, ethnicity and occupation, with 3,310,000 face images.
3）Wider Face
Released in 2018, it is 10 times larger than existing datasets. The dataset contains rich annotations including occlusions, poses, event classes, and facial bounding boxes. Faces in the proposed dataset are extremely challenging due to large variations in size, pose, and occlusion, with 32,203 face images.
Based on the security video surveillance system EasyCVR and AI video intelligent analysis system, integrating AI, cloud computing, big data and other technologies, it can identify people, objects, behaviors, etc. in the surveillance scene and issue alarms for abnormal situations, which can greatly satisfy The industry's smart supervision needs such as data sensing, intelligent detection, intelligent analysis, and intelligent alarms based on video services. Interested users can contact us or go to the demo platform to test and use it.
Bun releases official version 1.0, a magical bug in Windows File Explorer when JavaScript is runtime written by Zig , improves performance in one second JetBrains releases Rust IDE: RustRover PHP latest statistics: market share exceeds 70%, the king of CMS transplants Python programs To Mojo, the performance is increased by 250 times and the speed is faster than C. The performance of .NET 8 is greatly improved, far ahead of .NET 7. Comparison of the three major runtimes of JS: Deno, Bun and Node.js Visual Studio Code 1.82 NetEase Fuxi responded to employees "due to BUG was threatened by HR and passed away. The Unity engine will charge based on the number of game installations (runtime fee) starting next year.