Raspberry Pi and Sony have partnered to introduce a new AI camera module priced at $70, aimed at simplifying the development of edge AI visual data processing solutions. This Raspberry Pi AI camera integrates Sony's IMX500 image sensor and an onboard AI accelerator, allowing users to run neural network models directly on the device without the need for additional hardware such as GPUs or external accelerators.

The AI camera is centered around Sony's IMX500 intelligent vision sensor and is equipped with Raspberry Pi's RP2040 microcontroller for managing neural networks and firmware. This combination enables real-time image processing and inference, freeing up the Raspberry Pi processor to handle other tasks simultaneously.

Key Specifications of the AI Camera:

• Resolution: Up to 4056×3040 pixels at 10 frames per second; or 2028×1520 pixels at 30 frames per second
• Pixel Size: 1.55μm × 1.55μm
• Field of View: 78 degrees, with manual focus support
• Compatibility: Supports all Raspberry Pi models, including Raspberry Pi Zero

A notable feature of the AI camera is its seamless integration with Raspberry Pi's existing software stack. Utilizing the libcamera framework, the device allows developers to easily access synchronized image frames and neural network outputs. This integration simplifies the development of applications such as object detection and pose estimation, significantly reducing the amount of required code. For example, using libraries like Picamera2, developers can run common models like MobileNet SSD for object detection or PoseNet for pose estimation with minimal coding.

Additionally, the AI camera supports converting existing neural network models from popular frameworks like TensorFlow or PyTorch into formats that can efficiently run on the device. Sony's AI tool suite facilitates this conversion process and offers a model library with optimized models for users who wish to get started quickly.

While the AI camera provides significant advantages, Raspberry Pi has also launched a more powerful AI suite in June, delivering computational capabilities of up to 130 trillion operations per second. However, the AI suite is only compatible with Raspberry Pi 5 and requires a separate camera module. The choice between the AI camera and the AI suite depends on specific project requirements, such as performance needs and hardware compatibility.

The introduction of the AI camera opens up new possibilities for developers interested in edge AI applications. Potential use cases include smart city sensors for traffic monitoring and industrial quality assurance systems. The device's compact size (25×24×11.9 mm) and compatibility with all Raspberry Pi models make it a versatile tool for both hobbyists and professionals.

Raspberry Pi has released a getting started guide that includes instructions for hardware installation, software setup, and running example models from the model library, providing a reference for users interested in exploring the AI camera's capabilities.

The Raspberry Pi AI camera is now available for purchase, with the company committing to continue production until at least January 2028.