Face Recognition Engine(Route 16)FS1016
| Product function | Face capture, 1: N recognition, face retrieval, dynamic layout control, data statistics, black and white list multi-level management, controller management, equipment management, area management, real-time alert, etc.; Recognition ability: 1:1, 1:N Recognition accuracy, fast recognition, face tracking and detection time of about 20ms, face feature extraction time of about 200ms, face ratio time of about 0.1 MS |
| Product characteristics | Compact structure, easy to install, Wired, wireless, 232 output, can be connected, can be used as an external alarm or switch from the management ability, open secondary development interface. Facial critical point analysis module: according to the deep learning technology to locate the facial critical point, to realize feature vector extraction. Face standardization module: eliminates the influence of external environment such as external illumination, face posture, expression, and occlusion on the subsequent feature expression module. Feature expression module: Based on depth learning technology, different levels of face features are extracted. Metrics learning module: based on deep learning technology, different face features are mapped to a low-dimensional space. The difference of face features in low-dimensional space increases and the retrieval time decreases. Image feature search module: the index of human image feature, based on human image feature index for large-scale human image feature comparison. Structured information search module: generate spatial and temporal information, human image attribute information and other structural information indexes, based on structured information indexing for large-scale human image retrieval. Can access the mainstream protocol camera for capture. |
| Equipment performance | The number of single image multiple face captures reaches 30, the face pitch angle is less than 30 degrees, the left and right turns are less than 45 degrees, and the yaw angle is less than 30 degrees. Detection rate & GT; = 96 %, error rate = & lt; 0.1 %; Age: pupil spacing more than 15 pixels can be detected, pupil spacing more than 30 pixels can be recognized; Face detection and recognition speed: the length of face tracking and detection is not more than 20 MS, the length of face feature extraction is not more than 200 MS, and the face length is about 0.1 MS. Supports 30 * 30 to 3000 * 3000 pixel image input; Support single 720P/1080P access, support mainstream brand camera access, USB camera access; E5 processor, 64G memory, independent GPU graphics card, multiple cascades, load balancing, 100,000 face library |
| Applicable Object | Hotels, Internet cafes, venues, shopping malls, parks, communities, campuses, security, places involved in confidentiality and other final customers; An integrator that needs to integrate face recognition solutions |
