Firefly dual-lens AI Camera Module Supports Computer Vision at Edge
Firefly has launched two dual-lens AI camera modules with a built-in NPU (Neural Processing Unit) accelerator supporting up to 2 TOPS of performance. The modules feature Rockchip’s AI vision processors, RV1109 for CAM-C1109S2U module and RV1126 for CAM-C1126S2U module. The processor also supports FastBoot, TrustZone technology, and multiple crypto engines.
CAM-C1109S2U module’s RV1109 SoC comes with a 1.5GHz dual-core Arm Cortex-A7 CPU and a 400MHz RISC-V processor. It is also equipped with a 2D-capable graphics processor, and a VPU (Vision Processing Unit) compatible with H.264/H.265 encoding at 30 fps, and an NPU performing up to 1.2 TOPS.
CAM-C1126S2U module comes with a more powerful Rockchip RV1126 which has a quad-core Arm Cortex-A7 instead of the dual-core Arm Cortex-A7 in the CAM-C1109S2U module. Hence, this enhances the capacity of VPU encoding and decoding to allow up to 4K from the visible-light camera module and increases the NPU performance up to 2 TOPS.
Both the firefly camera modules have almost similar specifications, apart from the processors.
Technical Specifications of Firefly Camera Modules
- 1 or 2GB DDR4 RAM constitutes the system memory
- Storage of 8 or 16 GB eMMC 4.51 flash
- RGB camera features 1920×1080 resolution GC2093 image sensor with 650nm filter and 80cm focus distance
- IR camera features 1920×1080 resolution GC2053 image sensor with 850nm filter and 80cm focus distance
- 30-pin FPC header to connect MIPI display (TBC) for video output
- Working temperature ranges from 10 to 60°C
- Power Supply of 5V through USB Type-C port
- 84 x 22.45 x 19.35 mm dimensions with aluminum alloy housing.
The on-board video CODEC of the firefly camera modules supports 2K H.254/H.265 at 30FPS and multi-channel video encoding and decoding. Hence, it fulfills the requirement of low bit rate and low-latency encoding. Thus, making the video tenure smaller. The built-in 3F-HDR ISP and 3F-HDR feature multi-level noise reductions. It also comes with a dual-lens 2M (RGB+IR) WDR camera for working in strong light, backlight, and darkness.
The Firefly camera modules are compatible with the Linux environment. It operates with the Rockchip RKNN toolkit working in Windows and Mac OS. The AI camera module has dynamic connectivity for host mediums on android devices. The modules run on tools for AI algorithms for direct conversion and deployment of TensorFlow, PyTorch, Caffe, MxNet, DarkNet, ONNX frameworks. Hence, the firefly camera modules are suitable for various ML/DL applications. This includes face recognition and gesture recognition. It can also be used in domains like gate access control, smart security, smart IP camera, smart doorbell/peephole, self-service terminals, smart finance, smart construction, smart travel, and other industries.
For more information visit the official product page. CAM-C1109S2U and CAM-C1126S2U AI camera modules are available now for respectively $89 ad $95 on Firefly store. Images and technical specifications have also been taken from the product page.