TSBE3000 Face Server-Fornitore globale

3rd Floor, Building A, Aerospace Micro motor Factory, No. 25 Keji North 2nd Road, Songpingshan Community, Xili Street, Nanshan District, Shenzhen

The TSBE3000 facial recognition server integrates a BITAMANIN latest tensor processor BM1682 internally. 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 dual precision peak computing power, while the chip has up to 16MB of on-chip SRAM, which can greatly reduce data handling during model computation, improve performance, and reduce power consumption.
This system is suitable for applications such as facial gates, facial access control, facial attendance machines, and backend facial comparison analysis of cameras. It can receive captured photo streams, perform a 1:N facial comparison, and provide feedback on the comparison results to the gate and access control system, control the opening and closing of the gate and door, and synchronize the relevant results to the backend management system

Product features:
1. Perform inference calculations at the local edge rather than in the cloud
2. Enterprise level facial detection and recognition
3. Supports 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 backend management system, providing a basic public cloud management system
6. Provide SDK documentation for customers to develop upper level applications
7. Built in Bitmain BM1682 artificial intelligence processor, which can easily and quickly deploy existing mainstream or other trained neural network models such as DNN/CNN/RNN/LSTM.
Target function
Facial detection, facial recognition, and expression feature analysis, such as age, gender, etc
Facial tracking, deduplication, and image cutout
Facial recognition information reporting
Blacklist alarm
VIP account management

人脸识别参数

Usage scenarios	indoor
CPU/DL Accelerator	BM1682（4nucleusA53,3TFLOPSComputing power
Memory	8GB
Emmc	32GB
External expansion storage	TF card (maximum)256GB）
network interface	6 gigabit devices
power supply	220~230V
Whole machine power consumption	Average/60W
; Peak/90W	Heat dissipation methodHeat sink+Fan（PWM
Speed control)	-10^{working temperature}o^{~ +55}
o	storage
Local and cloud storage	working temperature
-10°-40℃	-40^{Storage temperature}o^{~ +70}
o	Working humidity
10%	~90%, no frost
transmission speed	10Mbps
100Mbps 1000Mbps	Switch Type
Gigabit switch	Number of interfaces6 of themNetwork standardsaccord withIEEE802.1D、
IEEE 802.1w	、IEEE 802.1sstandard
Exchange method	storage
-	forward
backplane bandwidth	32Gbps
MAC Address Table	12K
Facial detection rate	>99% (for video streaming)
Identification accuracy	>99%Misidentification rate<0.01%Facial detectable angle^{Rolling angle（}roll^） = -60o~ +60o^{Pitch angle（}pitch^） = -60^o~ +60^o
Yaw angle=-60	o~ +60oFacial recognition angle^{Yaw angle（}yaw^） = -30o~ +30o^{Pitch angle（}pitch^） = -30o~ +30o^{Rolling angle（}roll^）
= -30	o~ +30o
Face capture size	Face pixels are larger than4040*40
pixel	Facial recognition size20The distance between the eyes is greater thanpixelProcessing performance60Image stream:A picture/ Second or single image10Personal face
/	second
Video stream: simultaneous appearance in the frame	Zhang Face
Access capability	4 video streams or 10 image streams
Total capacity of the target database	100000 sheetsStorage image size<100KB (recommended)Storage image formatJPG，JPEG
，	PNG6，
BMP	Capture storage capacity6Video stream: ten thousand20Event storage capacity
Video stream:	ten thousand
Image stream:	ten thousand
OS (Operating System)	Debian GNU/Linux 9