Description
The Raspberry Pi AI Camera is a small module from Raspberry Pi that uses the Sony IMX500 Intelligent Vision Sensor. This sensor combines a 12-megapixel CMOS image sensor with built-in acceleration for neural network models, allowing users to create AI applications focused on vision without needing an external accelerator.
The AI Camera enhances still images or video by adding tensor metadata, while the main processor of the Raspberry Pi remains free for other tasks. With support for tensor metadata in the libcamera and Picamera2 libraries, as well as in the rpicam-apps suite, the camera is user-friendly for beginners while providing advanced users with extensive power and flexibility.
This camera works with all Raspberry Pi models and has the same PCB layout and mounting points as the Raspberry Pi Camera Module 3, while the overall depth is greater to accommodate the larger IMX500 sensor and optical subassembly
Key features:
- 12 MP Sony IMX500 Intelligent Vision Sensor
- Sensor modes: 4056x3040 at 10fps, 2028x1520 at 30fps
- 1.55 µm x 1.55 µm cell size
- 78-degree field of view with manually adjustable focus
- Integrated RP2040 for neural network and firmware management
The AI Camera is compatible with all Raspberry Pi models, including the Raspberry Pi Zero, and can be connected using standard camera ribbon cables by Raspberry Pi.
With Sony's AI tools, neural network models built on frameworks like TensorFlow or PyTorch can be adapted to run efficiently on the AI Camera. Alternatively, new models can be created to make use of the AI accelerator's specific capabilities.
Integration with Raspberry Pi libcamera
One of the main advantages of the AI Camera is its seamless integration with the Raspberry Pi camera software stack. Behind the scenes, libcamera processes the Bayer frame using Raspberry Pi's in-house ISP, just as it does for any other sensor.
Additionally, Raspberry Pi processes the neural network results to produce an output tensor, synchronizing it with the processed Bayer frame. Both the tensor and frame are then returned to the application during libcamera's request completion phase.
Raspberry Pi AI Kit or a Raspberry Pi AI Camera?
The AI Kit offers higher theoretical performance and supports a wider variety of models, but it ONLY works with the Raspberry Pi 5. In contrast, the AI Camera is smaller, more affordable if you don't already have a camera, and works with ALL Raspberry Pi models.
Both products deliver strong acceleration for common models and are optimized to integrate seamlessly with the Raspberry Pi camera software.
Obsolescence:
Raspberry Pi AI Camera will remain in production until at least January 2028.
Model Overview
Sensor | Sony IMX500 |
Resolution | 12.3 megapixels |
Sensor size | 7.857 mm (type 1/2.3) |
Pixel size | 1.55 µm × 1.55 µm |
Horizontal/vertical | 4056 × 3040 pixels |
IR cut filter | Integrated |
Autofocus system | Manual adjustable focus |
Focus range | 20 cm - ∞ |
Focal length | 4.74 mm |
Horizontal field of view | 66 ±3 degrees |
Vertical field of view | 52.3 ±3 degrees |
Focal ratio (F-stop) | 1.79 |
Infrared sensitive | No |
Output | Image (Bayer RAW10), ISP output (YUV/RGB), ROI, metadata |
Input tensor maximum size | 640(H) × 640(V) |
Input data type | 'int8' or 'uint8' |
Memory size | 8388480 bytes for firmware, network weight file, and working memory |
Framerate | 2×2 binned: 2028×1520 10-bit 30fps Full resolution: 4056×3040 10-bit 10fps |
Dimensions | 25 × 24 × 11.9 mm |
Ribbon cable length | 200 mm |
Cable connector | 15 × 1 mm FPC or 22 × 0.5 mm FPC |
Operating temperature | 0°C to 50°C |