Raspberry Pi AI Camera (IMX500): specs and pinout
A Raspberry Pi camera built on the Sony IMX500 Intelligent Vision Sensor, which combines a 12-megapixel sensor with on-chip neural network inference. It adds tensor metadata to the image stream so the host Raspberry Pi is free for other work.
Specs
| Spec | Value | Source |
|---|---|---|
| Sensor | Sony IMX500, 12.3 megapixel, 4056 x 3040 pixels, 7.857 mm (type 1/2.3), pixel size 1.55 um x 1.55 um | source Specification (page 3) |
| Frame rate | 2x2 binned 2028 x 1520 10-bit at 30 fps; full resolution 4056 x 3040 10-bit at 10 fps | source Specification (page 3) |
| Neural network input | Maximum input tensor 640 x 640; data type int8 or uint8; 8,388,480 bytes for firmware, network weights and working memory | source Specification (page 3) |
| Optics | Manual adjustable focus, 20 cm to infinity; focal length 4.74 mm; F1.79; horizontal FoV 66.3 +/- 3 degrees, vertical 52.3 +/- 3 degrees; IR cut filter integrated, not infrared sensitive | source Specification (page 3) |
| Output | Bayer RAW10 image, ISP output (YUV/RGB), ROI and metadata | source Specification (page 3) |
| Dimensions and cable | 25 x 24 x 11.9 mm; ribbon cable 200 mm; 15 x 1 mm FPC or 22 x 0.5 mm FPC connector | source Specification (page 3) |
| Operating temperature | 0 C to 50 C | source Specification (page 3) |
| Model firmware | Firmware and network files load onto the IMX500 at start; if not cached it can take several minutes (imx500-all package) | source Getting started > Install the IMX500 firmware |
Pinout
Rows are the 15-pin CSI connector on the Raspberry Pi side, as documented by Raspberry Pi; the camera board's own schematic is not covered for this variant. Raspberry Pi 5 and the Zero series use a 22-pin connector (different layout), so they need a different cable. The brief lists a 15 x 1 mm or 22 x 0.5 mm FPC for the camera.
| Pin | Function | Voltage | Notes | Source |
|---|---|---|---|---|
| 1 | GND | Ground | source Pinout information > Camera connector pinout (15-pin); host-side table | |
| 2 | CAM_DN0 | D-PHY lane 0 (negative), input | source Pinout information > Camera connector pinout (15-pin); host-side table | |
| 3 | CAM_DP0 | D-PHY lane 0 (positive), input | source Pinout information > Camera connector pinout (15-pin); host-side table | |
| 4 | GND | Ground | source Pinout information > Camera connector pinout (15-pin); host-side table | |
| 5 | CAM_DN1 | D-PHY lane 1 (negative), input | source Pinout information > Camera connector pinout (15-pin); host-side table | |
| 6 | CAM_DP1 | D-PHY lane 1 (positive), input | source Pinout information > Camera connector pinout (15-pin); host-side table | |
| 7 | GND | Ground | source Pinout information > Camera connector pinout (15-pin); host-side table | |
| 8 | CAM_CN | D-PHY clock (negative), input | source Pinout information > Camera connector pinout (15-pin); host-side table | |
| 9 | CAM_CP | D-PHY clock (positive), input | source Pinout information > Camera connector pinout (15-pin); host-side table | |
| 10 | GND | Ground | source Pinout information > Camera connector pinout (15-pin); host-side table | |
| 11 | CAM_IO0 | 3.3 V | GPIO, typically an active-high power enable; bidirectional | source Pinout information > Camera connector pinout (15-pin); host-side table |
| 12 | CAM_IO1 | 3.3 V | GPIO (for example clock or LED); some products do not include it; bidirectional | source Pinout information > Camera connector pinout (15-pin); host-side table |
| 13 | SCL | 3.3 V | I2C clock; pulled up to 3.3 V on the Raspberry Pi board | source Pinout information > Camera connector pinout (15-pin); host-side table |
| 14 | SDA | 3.3 V | I2C data; pulled up to 3.3 V on the Raspberry Pi board | source Pinout information > Camera connector pinout (15-pin); host-side table |
| 15 | 3V3 | 3.3 V | 3.3 V supply, output from the Raspberry Pi | source Pinout information > Camera connector pinout (15-pin); host-side table |
Build with it
Project ideas for the Raspberry Pi AI Camera (IMX500). Each link opens Soldr with the prompt filled in; Soldr assigns the pins, checks power and shows how to wire every part.
- Pet-visit counter: The IMX500 runs the detection network on the camera, so the Pi 4 only reads tensor metadata. Learning project; if people can be in view, tell them the camera is running.
- Workbench tool detector: Raspberry Pi's documentation shows detection overlays through rpicam-apps. Focus is manual and starts at 20 cm, so set the table distance.
- Pose-based exercise counter for yourself: The AI Camera documentation runs a pre-packaged PoseNet model. Learning project, not a coaching or medical tool; only film yourself or people who agreed.
- Custom model deployment test: The documentation covers converting and packaging a model for the camera (input up to 640 x 640, int8/uint8). The Pi 5 needs a 22-pin camera cable (not included).
Common mistakes
- Expecting a model to run the moment the camera is plugged in: install imx500-all and reboot; the first start loads firmware and can take minutes (source: link, Getting started).
- Using a 15-pin cable on a Raspberry Pi 5 or Zero: they use a 22-pin camera connector, and the brief lists both 15 x 1 mm and 22 x 0.5 mm cable options (source: link, page 3).
- Hot-plugging: the brief says to shut down and disconnect external power before connecting the device (source: link, Safety instructions).
Not covered by the official source
- No current figure is published in the brief or documentation.
- The pin table is the 15-pin host connector; the camera board schematic is not covered.