Product overview
SenseNets Technology Ltd. (“SenseNets”), a NetPosa-controlled company that engages in application of AI technologies in the security and protection field, is committed to in-depth exploration of the value of videos, and focuses on providing feasible solutions for users. The series products of portrait analysis server, which employ the face recognition algorithm based on deep learning, are used by various industries such as public security, healthcare, finance and education. Covering core functions including dynamic deployment and control and library retrieval, such products support both stand-alone and clustering architecture, and represent standard products of high performance with access provided for interfaces of basic application system.
The portrait analysis server can be divided into face deployment & control server and static library server in terms of types, and engine and all-in-one machine in terms of product form. Among these, the engine supports stand-alone and clustering modes, and can be used on big data platforms; and the all-in-one machine support operation of multiple servers while containing a basic application system.
Face deployment & control server
System composition
Main functions
1. Real-time deployment & control
Information such as face images, features and attributes collected from the videos shot by cameras will form structured data in real time; and the real-time snapshot of faces by monitors will be compared with the personnel in the dynamic deployment & control library, and alarm will be triggered immediately in case of matching.
2. Alarm management
The historical alarm records are searched by inputting limited conditions such as name, ID card No, similarity scope, deployment & control library and time, to view the alarm details.
3. Passage records
It supports searching historical snapshot information through screening face attributes or uploading face images.
4. Retrieval records
The historical retrieval records of the system can be consulted, to facilitate analysis based on case combination.
5. Statistical analysis
The passage flow of personnel and alarm quantity will be counted to directly display the data distribution.
Product features
Designed based on GPU operation, it has strong performance, and the single card is able to process 8-channel HD real-time flows concurrently;
With application of deep learning algorithm, it features high recognition accuracy and high adaptability to scenarios;
Supporting stand-alone or clustering architectural method, it can be extended flexibly based on actual needs;
The all-in-one machine, which integrates the functions of web system on the server, removes tedious installation and deployment, thus can be directly applied in various operations;
Providing open outward call interface, it can be invoked by third-party platforms and support second-system effect.
Application scenarios
Software specification
Hardware specification & system performance indicators
Static server with large storage
Functional architecture
1. Face retrieval
It supports uploading images of personnel, which will be compared with the images in the face library, and information of the personnel in the library with higher similarity will be returned; it allows face retrieval by various conditions such as the intervals of gender, ethnic group and age; as well as allows uploading two face images for comparison and outputs the face similarity value.
2. Retrieval records
The records of face retrieval, ID retrieval and face library retrieval can be consulted, to carry out analysis by integrating information of more cases.
3. Statistical analysis
The data statistics diagram of the face library in the system can be viewed.
4. Face library management
The face library of the system can be created and managed, to support single and batch import of face information.
5. User management
The users of the system can be managed, including actions such as “create”, “revise”, “delete”, “disable” and “enable”, etc.
6. Service status
The service operation status of the system can be viewed, to troubleshoot the system operation.
Product features
Designed based on GPU operation, it has strong performance, and the single machine is able to support 8-million-capacity static library;
With application of deep learning algorithm, it features high recognition accuracy and high adaptability to scenarios;
Supporting stand-alone or clustering architectural method, it can be extended flexibly based on actual needs;
The all-in-one machine, which integrates the functions of web system on the server, removes tedious installation and deployment, thus can be directly applied in various operations;
Providing open outward call interface, it can be invoked by third-party platforms and support second-system effect.
Application scenarios
1. The face image of a suspect can be acquired through clips of the monitoring videos and on-site filming, which will be subject to 1:N comparison and retrieval with the permanent population library and the escaping convict library, so that the information of the personnel in the library that has a highly similar face will be obtained, and the public security bureau or other users will be able to judge the identity of the suspect more easily;
2. The similarity of faces in the two images will be analyzed, to help judging whether they are the same person.
Product diagram
Product specification
