Raspberry Pi AI Camera Module with Integrated Machine Learning

Raspberry Pi has released its latest camera module, the Raspberry Pi AI Camera, designed to bring on-device machine learning capabilities directly to users. Built around the Sony IMX500 Intelligent Vision Sensor, the AI Camera integrates a 12.3-megapixel CMOS sensor with an onboard AI accelerator. This design allows the camera to perform inferencing tasks, such as object detection and pose estimation, while capturing images or video, reducing the processing load on the Raspberry Pi’s CPU.
  I already showed you a sneak peek of this not-very-well-kept secret at Embedded World 2024 earlier this year, and now they’re officially shipping.
   

Raspberry Pi AI Camera Key Specifications

The camera uses Sony’s IMX500 sensor, which features a 1/2.3-inch sensor size with 1.55 μm pixels and a resolution of 4,056×3,040 pixels. The camera supports both Bayer RAW10 and ISP output formats (YUV/RGB) and includes support for Region of Interest (ROI) cropping. For image processing, it outputs tensor metadata, providing real-time data that AI applications can use.
 

Notable specifications include:

• Focal Length: 4.74 mm
• Field of View: Horizontal 66° ±3°, Vertical 52.3° ±3°
• Focus Range: 20 cm to ∞, with manual adjustable focus
• Framerate: 30 fps in 2×2 binned mode at 2,028×1,520 resolution; 10 fps at full 4,056×3,040 resolution
• Physical Dimensions: 25×24×11.9 mm
• Operating Temperature: 0°C to 50°C

The Raspberry Pi AI Camera’s footprint and mounting locations are identical to the Raspberry Pi Camera Module 3, but it is slightly deeper to accommodate the larger sensor and optical components​.

Software Integration and AI Capabilities

The AI Camera’s integration with Raspberry Pi’s software ecosystem is comprehensive. Tensor metadata is supported natively by the libcamera and Picamera2 libraries, allowing developers to implement AI features in their applications easily. The camera also works with the rpicam-apps software suite, simplifying AI deployment for less experienced users.

Setting up the AI Camera involves installing a set of software packages, including firmware and models for the IMX500 sensor. Once configured, users can run pre-installed demos, such as Mobilenet SSD for object detection or Posenet for pose estimation. For developers aiming to train custom models, Raspberry Pi provides tools to convert models for use with the IMX500, supporting both PyTorch and TensorFlow environments​.

Applications and Use Cases

The AI Camera opens up new possibilities for vision-based AI applications, including robotics, automation, and security systems. Its on-device AI processing reduces latency and offloads the computational burden from the Raspberry Pi itself, making it a suitable choice for projects requiring real-time inferencing. The camera is expected to remain available through at least January 2028​.

The Raspberry Pi AI Camera, with its combination of high-resolution image capture and onboard inferencing, provides a practical solution for users looking to integrate machine learning into their projects without needing additional hardware accelerators. Its compatibility with all Raspberry Pi models and the robust software support makes it an appealing option for both beginners and advanced users working on AI-driven applications.