The Raspberry Pi AI Kit bundles the Raspberry Pi M.2 HAT+ with a Hailo AI acceleration module for use with Raspberry Pi 5. It provides an accessible, cost-effective, and power efficient way to integrate high-performance AI. Explore applications including process control, security, home automation, and robotics.

Kit Contains:

1. Hailo AI Module:

• Neural Processing Unit (NPU):
  • Chipset: Hailo-8L
  • Performance: 13 TOPS (Tera-Operations per Second). Capable of delivering high-performance inference acceleration for neural networks such as CNNs (Convolutional Neural Networks) and RNNs (Recurrent Neural Networks).
• Form Factor:
  • M.2 2242 (22 mm width, 42 mm length), compact design optimized for integration into embedded systems like Raspberry Pi.
• Interface:
  • M Key Edge Connector for M.2 slots.
  • PCIe Gen 2.0 x1 lane connection to the host (Raspberry Pi 5), allowing a data bandwidth of up to 5 GT/s (Giga-transfers per second).
• Power Efficiency:
  • Low power consumption for AI inference tasks, optimizing energy usage in applications such as robotics and IoT.
• Temperature Management:
  • Thermal pad pre-fitted between the module and the M.2 HAT+ to dissipate heat during intensive processing.
• Power Requirements:
  • Supply Voltage: 5V (powered via the M.2 HAT+ from the Raspberry Pi).

2. Raspberry Pi M.2 HAT+:

• Form Factor:
  • Designed to fit on top of the Raspberry Pi 5.
• Connectivity:
  • M.2 Key M Slot for the Hailo AI module.
  • PCIe 2.0 Interface to communicate with the Raspberry Pi 5.
• GPIO Header Compatibility:
  • Includes a 16mm stacking GPIO header that raises the HAT+ to provide clearance for additional accessories and GPIO connections.
• Thermal Management:
  • Thermal pad pre-installed for passive cooling of the AI module.
• Power Delivery:
  • The HAT+ delivers power to the AI module through the PCIe interface, simplifying power management.
• Mounting Hardware:
  • Kit includes spacers, screws, and washers for secure mounting of the HAT+ to the Raspberry Pi.

Technical Specifications:

AI Module (Hailo-8L):

• Chipset: Hailo-8L Neural Processing Unit (NPU)
• Performance: 13 TOPS AI inference
• Form Factor: M.2 2242
• Interface: PCIe Gen 2.0 x1 (M.2 Key M slot)
• Power Consumption: Optimized for energy-efficient AI processing
• Temperature Range:
  • Operating Temperature: 0°C to 70°C
  • Storage Temperature: -40°C to 85°C
• Heat Dissipation: Thermal pad included for passive cooling
• Supported Frameworks:
  • TensorFlow Lite
  • ONNX (Open Neural Network Exchange)
  • Caffe and PyTorch models (via Hailo SDK)
• Supported Neural Network Architectures:
  • CNN (Convolutional Neural Networks)
  • RNN (Recurrent Neural Networks)
  • MLP (Multi-Layer Perceptron)
  • SSD (Single Shot MultiBox Detector)
  • Yolo (You Only Look Once)

Raspberry Pi M.2 HAT+:

• Interface: PCIe Gen 2.0 x1
• Compatible Raspberry Pi Model: Raspberry Pi 5
• Form Factor: Full-size HAT for Raspberry Pi
• Power: Delivered through the PCIe interface from Raspberry Pi
• Thermal Management: Pre-installed thermal pad for heat dissipation
• Physical Connections:
  • M.2 Key M slot for the AI module
  • GPIO headers (16mm stacking height for accessory compatibility)
• Hardware Compatibility:
  • Pre-configured to work with Raspberry Pi 5
• Mounting Hardware: Comes with screws and spacers for secure attachment
• Dimensions: Standard Raspberry Pi HAT size

The AI module is a 13 tera-operations per second (TOPS) neural network inference accelerator built around the Hailo-8L chip. The module uses the M.2 2242 form factor, and comes pre-installed in the M.2 HAT+, to which it connects through an M key edge connector. The M.2 HAT+ communicates between the AI module's M.2 interface and the Raspberry Pi 5's PCIe 2.0 interface. When the host Raspberry Pi 5 is running an up-to-date Raspberry Pi OS image, it automatically detects the Hailo module and makes the NPU available for AI computing tasks. The built-in rpicam-apps camera applications in Raspberry Pi OS natively support the AI module, automatically using the NPU to run compatible post-processing tasks.