Google Coral Edge TPU ASIC M.2 A+E key accelerator, coprocessor for Frigate

Google Coral Edge TPU ASIC M.2 A+E key accelerator, coprocessor for Frigate

The Coral M.2 Accelerator A+E key module is an advanced device that enables integration of the Edge TPU from Google coprocessor with existing systems. It significantly improves performance of machine learning (ML) inference processes while maintaining low power consumption. This coprocessor can efficiently process advanced vision models such as MobileNet v2, reaching speeds of up to 400 frames per second.

It is designed as a compact card that fits universal M.2 slots. This makes it easy to add artificial intelligence capabilities to a wide range of products, including embedded platforms and mini computers (ideal e.g. for Raspberry Pi), especially those running Linux. The module supports a broad range of hardware platforms, including ARM and x86 architectures.

Key technical specifications

• Use case: Machine learning (ML) and artificial intelligence (AI)
• Hardware interface: M.2 A+E Key card (M.2-2230-A-E-S3)
• Bus: PCIe Gen2 x1
• System requirements: Any computer with Linux (Debian 6.0+ or derivatives, e.g. Ubuntu 10.0+)
• Architecture: x86-64 or ARM32/64 with ARMv8 instruction set
• Supply voltage: 3.3 V DC (±10% tolerance)
• Operating temperature: −20 to 70 °C
• Physical dimensions: 22 mm x 30 mm x 2.3 mm
• Weight: 3.1 g

Example applications

• Accelerating video surveillance systems (e.g. Frigate) in real time
• Object detection and classification in industrial applications
• Smart home systems based on image analysis
• Prototyping edge devices with artificial intelligence
• Advanced robotics requiring fast vision data processing

Benefits

• Large performance gain for ML model inference
• Very low power consumption relative to compute performance
• Easy integration thanks to the standard M.2 A+E key connector
• Broad compatibility with popular architectures (ARM, x86) and Linux systems
• Miniature size and low weight for mounting in compact devices