Raspberry Pi AI Camera – Edge Inference with Integrated Vision Sensor
The Raspberry Pi AI Camera features the Sony IMX500 Intelligent Vision Sensor, combining a 12.3MP CMOS sensor with built-in neural processing. It performs inference directly on the sensor, enabling computer vision tasks like detection and segmentation without needing an external accelerator. Tensor metadata can be processed via rpicam-apps, libcamera, or Picamera2.
Key Features
- Sensor: Sony IMX500 (Intelligent Vision Sensor)
- Resolution: 12.3MP (4056×3040), 1/2.3" sensor size
- Pixel size: 1.55µm × 1.55µm
- Output: RAW10, YUV/RGB, ROI, tensor metadata
- Inference input: Max 640×640, int8 or uint8
- Manual focus lens, 4.74mm focal length, F1.79
- Field of view: 66° horizontal, 52° vertical
- Framerate: Up to 30fps (binned), 10fps (full res)
- Memory: 8.4MB (firmware, weights, working buffer)
- Dimensions: 25×24×11.9mm
- 200mm ribbon cable, supports 15-pin or 22-pin CSI
- Compatible with all Raspberry Pi models
- Production guaranteed until at least January 2028
Note: Inference runs directly on the sensor. Host CPU remains free. Not IR sensitive.
Product safety information
Manufacturer information:
Raspberry Pi Ltd.
Cambridge Science Park, Milton Road 194
Cambridge, Vereinigtes Königreich, CB4 0AB
compliance@raspberrypi.com
https://www.raspberrypi.com