Specifications
Sensor
Sony IMX500 Intelligent Vision Sensor with integrated logic
Resolution
12.3 MP, 4056 x 3040 active pixels
Pixel size
1.55 µm x 1.55 µm
Video mode
1080p30 on Raspberry Pi hosts
Binned mode
2x2 binning at 2028 x 1520, 10-bit
Shutter type
Rolling shutter, colour
Camera interface
SPI and MIPI CSI
Lens mount
M12
Field of view
82° diagonal x 72° horizontal x 56° vertical
SKU B0642-1
Camera module for MCU platforms, 12.3 MP AI camera
SKU B0642-2
ESP32-P4 AI camera bundle kit, AI demo at 1024 x 600 and 20 fps
Supported hosts
ESP32-P4, Raspberry Pi, Raspberry Pi Pico 2, Arduino and other MCU/SoC boards
Inference location
On-sensor — models run in the IMX500 itself
Host requirement
No Linux stack required for inference workloads
SDK
ArduCam imx500-mcu-sdk with ESP32-P4, Raspberry Pi and Pico 2 examples
Model support
IMX500 AI model database with ready label sets
Operating profile
Low latency, low power, privacy-preserving local processing
Overview
Arducam's IMX500 series turns Sony's IMX500 intelligent vision sensor into a general-purpose embedded camera. The IMX500 stacks pixel sensing and logic computation on a single die, so inference happens inside the sensor rather than on the host — collapsing latency, removing cloud dependency and keeping raw imagery on device.
The series is deliberately host-agnostic. Instead of assuming a Linux-capable board, it exposes both SPI and MIPI CSI interfaces and ships SDK examples for ESP32-P4, Raspberry Pi and Raspberry Pi Pico 2, with support extending to Arduino-class hardware. A 12.3 MP colour sensor at 4056 x 3040 feeds a M12 lens mount with an 82° diagonal field of view, delivering 1080p30 video on Raspberry Pi hosts and 2x2 binned 2028 x 1520 10-bit output when full resolution is not required.
Two SKUs cover most designs: B0642-1, the bare module for MCU platforms, and B0642-2, an ESP32-P4 bundle kit that runs a visible AI demo at 1024 x 600 and 20 fps out of the box. Arducam maintains an accompanying model label database so deployed demos move to production models without re-tooling the pipeline.
Key Benefits
Inference at the sensor — networks execute on the IMX500 itself, so systems avoid shipping frames and avoid a Linux host entirely. MCU-friendly interfaces — SPI as well as MIPI CSI widens host choice to ESP32-P4, Pico 2 and Arduino boards. Privacy by architecture — image data can stay on device with only results leaving the module. Fast path to a demo — the ESP32-P4 bundle kit ships with a working on-screen AI demo and an accompanying SDK and label database.
Applications
MCU-class machine vision for industrial automation and safety systems, forklift and vehicle collision avoidance, presence and people counting without cloud upload, battery-powered vision sensors where power budget rules, and embedded inspection tasks needing deterministic on-device inference.
Request a Quote — ARDUCAM IMX500 AI CAMERA — 12.3MP ON-SENSOR INFERENCE MODULE
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →