Raspberry Pi has launched an innovative AI Camera Module, marking a significant advancement in on-device artificial intelligence for computer vision applications. Priced at £70, this new offering combines a high-quality image sensor with on-board AI processing capabilities, opening up a world of possibilities for developers and businesses alike.
At the heart of the AI Camera Module is Sony's IMX500 'Intelligent Vision Sensor', a 12.3-megapixel CMOS image sensor with integrated AI processing. This powerful combination allows for real-time object detection, segmentation, and pose estimation directly on the camera, freeing up the Raspberry Pi's CPU and GPU for other tasks.
The module's specifications are impressive. It captures images at a resolution of 4,056 x 3,040 pixels, with frame rates of 10 fps at full resolution or 30 fps when binned to 2,028 x 1,520. The camera features a manual focus lens with a 20cm to infinity range and a 66-degree horizontal field of view.
One of the most remarkable aspects of the AI Camera Module is its ability to run complex neural networks in real-time. Pre-installed models like MobileNet SSD for object detection and PoseNet for pose estimation can process 30 frames per second without taxing the host Raspberry Pi's resources. This is achieved through 8MB of on-device RAM, shared between firmware, model weights, and working memory.
Raspberry Pi has also launched a 'model zoo', providing Python examples for image classification, object detection, and segmentation. Supported models include EfficientNET v2, MobileViT XS, ResNet18, SqueezeNet, and YOLOv8n, with varying input resolutions to accommodate different performance needs.
The AI Camera Module is compatible with all Raspberry Pi computers, making it a versatile addition to existing setups. It's particularly well-suited for industrial and embedded applications, which now account for over 70% of Raspberry Pi's sales.
Potential use cases for the AI Camera Module are diverse. In smart cities, it could detect empty parking spots or monitor traffic flow. Industrial environments might employ it for automated quality assurance. The module's compact size and powerful capabilities make it ideal for edge AI applications where processing visual data efficiently is crucial.
While the AI Camera Module is nearly three times the price of the standard Raspberry Pi Camera Module 3, its on-board AI processing justifies the cost for many applications. However, users primarily interested in basic video streaming or still-image capture may find the Camera Module 3 more suitable, especially for battery-powered projects.
Raspberry Pi's commitment to long-term production, promising availability until at least January 2028, makes the AI Camera Module an attractive option for companies looking to integrate vision-based AI into their products or processes. This new offering further cements Raspberry Pi's position as a key player in the world of accessible, powerful computing solutions for both hobbyists and industry professionals.
