Taiwan based iCatch Technology has recently announced iCatch V37 SoC for AI camera applications. The processor can capture 4Kp30 video thanks to the company’s 7th generation image signal processor (ISP) and H.264 high-performance encoder, and includes a neural processing unit (NPU) to enable features such as people detection.
The V37 series includes V37A and V37M, where V37M has an embedded DRAM in the package to enable smaller form factor designs.
iCatch V37 4K AI camera SoC key features:
- 32-bit RISC CPU
- Dual CIS Support
- HDR Combination
- 5-axis Electronic Image Stabilization
- Dewarping Engine
- Video Noise Reduction (MCTF)
- Rolling Shutter Compensation
- H.264 4Kp30 Video Encode
- Neural Processing Unit
- Security Engine
- USB 2.0, HDMI v1.4b, GMAC
- External WiFi/BT Support
The SoC can be used to design products such as digital still camera (DSC), action camera, 360° VR Camera, digital video recorder (DVR), home security camera, or body worn camera.
One example of use case for the NPU is home security camera applications, where the camera can recognize family members and only send out alerts when strangers or unknown pets enter in the scene. The NPU can also be used for scene recognition, voice commands, human tracking and more. The 5-axis EIS is said to remove the need for a gimbal, as the camera can handle image stabilization on its own just as well.
V37 processor is their first product with an NPU, but not their only product, as they also released a dual core, quad camera version (V50) last year, and two camera SoCs for the automotive market are also offered. There are not that many details on the company’s website, but a person not affiliated with the company told me their products are quite price competitive, so if you have a (4K) camera project, it may be worth investigating their processors and solutions.