WDR Dual-Lens Camera Module, equipped with high-performance AI intelligent vision processor, has powerful AI computing performance, supports various AI frameworks and human detection with high recognition accuracy. It is suitable for face recognition, gate access control, gesture recognition, expression recognition, face attribute analysis, etc. Abundant resources facilitate secondary development and help the project to be implemented quickly.
Low-consumption AI vision processor RV1109, with 14nm lithography process and dual-core 32-bit ARM Cortex-A7 architecture, integrates NEON and FPU — the frequency is up to 1.5GHz. It supports FastBoot, TrustZone technology and multiple crypto engines.
Built-in neural network processor NPU with computing power up to 1.2 Tops realizes that the power consumption of AI computing is less than 10% of the power required by the GPU. With tools and supporting AI algorithms provided, it supports direct conversion and deployment of Tensorflow, PyTorch, Caffe, MxNet, DarkNet, ONNX, etc.
With built-in 3F-HDR ISP, multi-level noise reduction, 3F-HDR, and other technologies, it is equipped with dual-lens 2M (RGB+IR) WDR camera; it not only meets the scenes of strong light, backlight and darkness, but also realizes human detection and anti-spoofing functions of face recognition.
The module provides face recognition algorithms and supports face database up to 100,000 capacity; mask-on face recognition, human detection and face comparison are supported — with efficient recognition speed the recognition accuracy rate over 99.7%.
Built-in Video CODEC supports 2K H.254/H.265@30FPS and multi-channel video encoding and decoding, meeting the needs of low bit rate, low-latency encoding, perceptual encoding and making the video occupancy smaller.
It can be used as a USB device to output visual processing data and images to the computer; or can be used as an IPCAM monitoring and identification front end to connect to the NVR through the network port; and also can be connected to a 1080P screen through MIPI to make a face recognition or an AI vision terminal.
Equipped with all-aluminum alloy shell, the space-saving module has efficient heat dissipation and can be flexibly embedded in various AI products.
A complete SDK, including cross compiler toolchain, BSP source code, application development environment, development documents, examples and other resources, is provided for the users to make a further customization.
It is widely used in face recognition, gesture recognition, gate access control, smart security, smart doorbell/peephole, self-service terminals, smart finance, smart construction, smart travel and other industries.
CAM-C1109S2U | CAM-C1126S2U | |
SoC |
RV1109 |
RV1126 |
CPU |
Dual-core ARM Cortex-A7 32-bit , frequency 1.5GHz, integrate NEON and FPU Each core has a 32KB I-cache, 32KB D-cache and 512KB shared second-level cache Based on RISC-V MCU |
Quad-core ARM Cortex-A7 32-bit, frequency 1.5GHz, integrate NEON and FPU Each core has a 32KB I-cache, 32KB D-cache and 512KB shared second-level cache Based on RISC-V MCU |
NPU |
1.2Tops, support INT8/ INT16, It has strong network model compatibility and provides RKNN tool, which can realize the conversion of commonly used AI framework models Eg: (caffe, darknet, mxnet, onnx, pytorch, tensorflow, tflite) and algorithm support |
Computing power up to 2.0Tops, supports INT8/ INT16, It has strong network model compatibility and provides RKNN tool, which can realize the conversion of commonly used AI framework models Eg: (caffe, darknet, mxnet, onnx, pytorch, tensorflow, tflite) and algorithm support |
RAM |
1GB / 2GB DDR3 |
1GB / 2GB DDR4 |
存储 |
8GB / 16GB eMMC |
8GB / 16GB eMMC |
Video Encoding |
H.264/H.265 encoding capability: 2688 x 1520@30 fps+1280 x 720@30 fps 3072 x 1728@30 fps+1280 x 720@30 fps 2688 x 1944@30 fps+1280 x 720@30fps |
4K H.264/H.265 30fps video encoding: 3840 x 2160@30 fps+720p@30 fps encoding |
Video Decoding |
5M H.264/H.265 decoding |
4K H.264/H.265 30fps video decoding 3840 x 2160@30 + 3840 x 2160@30 fps decoding |
OS |
Linux |
Linux |
Power |
5V |
5V |
Operating Temperature |
-10℃~60℃ |
-10℃~60℃ |
Operating Humidity |
10%~90 % |
10%~90 % |
Dual-Lens Camera | ||
Camera (IR) | Camera (RGB) | |
Image Sensor |
GC2053 |
GC2093 |
Sensor Size |
1 / 2.9 |
1 / 2.9 |
Resolution |
1920*1080 |
1920*1080 |
Pixel Size |
2.8 μm |
2.8 μm |
Output Format |
RAW |
RAW |
Interface |
MIPI |
MIPI |
Focus Distance |
80 cm |
80 cm |
Lens |
4P |
4P |
Filter |
850 nm |
650 nm |
Field Angle |
D70°H62°V38° |
D70°H62°V38° |
Distortion |
≤0.5% |
≤0.5% |
Aperture/ Focal Length |
F2.0/4.3mm |
F2.0/4.3mm |
Resolving Power |
Center: 800 Around: 600 |
Center: 800 Around: 600 |
Face Recognition & Detection | ||
Max Database |
100,000 |
|
Recommended Database |
10,000 |
|
Accuracy |
Under standard test environment, 10,000 database No Mask: Accuracy 99% With mask: Accuracy 95% |
|
Face Detection |
Face detection time: ~23ms Face tracking time: ~7ms |
|
Human Detection |
Single-lens detection time: ~45ms Dual-lens detection time: ~15ms |
|
Face Recognition |
Feature extraction time: ~25ms Single recognition time: ~0.0115ms |
|
Recommended Image Input Size |
720P |
|
Min Face Recognition Size |
50*50 pixel (Non-human) 90*90 pixel (Human) |
|
Recommended Face Recognition Angle |
Yaw ≤ ±30° Pitch ≤ ±30° Roll≤ ±30° |
|
Appearance | ||
Dimension |
84.0mm × 22.45mm × 19.35mm |
|
Shell |
All aluminum alloy design, efficient heat dissipation |