tsbe3000 face recognition server integrates a newly released tensor processor bm1682 of bitamanin. each bm1682 has 64 npu processing units and each npu has 32 eu computing units. a single bm1682 chip can provide up to 3tflops of double-precision peak computing power. at the same time, the chip has up to 16mb on-chip sram, which can greatly reduce the data handling during model operation, improve performance and reduce power consumption.
the system is suitable for face gate, face access control, face attendance machine, camera back-end face comparison, and analysis. it can receive the captured photostream, and then conduct face 1: n comparison, and feedback the comparison results to the gate and access control, control the switch of the gate and door and synchronize the relevant results to the background management system
1. reasoning at local edge rather than cloud
2. enterprise level face detection and recognition
3. support 10 channels of jpeg photo input or 4 channels of dynamic rtsp video stream input
4. support liteos local management system
5. support synchronous detection and comparison results to the cloud background management system, and provide the basic public cloud management system
6. provide sdk documents for customers to develop upper application level
7. the built-in bm1682 artificial intelligence processor of bitcontinental can facilitate the rapid deployment of existing mainstream or other trained neural network models such as dnn / cnn / rnn / lstm.