Introducing the Coral USB Accelerator, a powerhouse for ML (machine learning) inferencing on existing Linux systems. Equipped with Google's Edge TPU, this small ASIC delivers high-performance ML inferencing at a minimal power cost through a USB 3.0 interface. With capabilities like executing MobileNet v2 at over 100 fps, it enables rapid ML inferencing on embedded AI devices while conserving power and preserving privacy.
Models are developed in TensorFlow Lite and compiled to run seamlessly on the USB Accelerator. Key benefits include high-speed inferencing, low power consumption, and a compact footprint. The Edge TPU supports TensorFlow Lite and is compatible with Google Cloud.
Specifications:
- Google Edge TPU ML accelerator coprocessor
- USB 3.0 Type-C socket
- Supports Debian Linux on host CPU
- Arm 32-bit Cortex-M0+ Microprocessor with up to 32 MHz max, 16 KB Flash memory with ECC, and 2 KB RAM
- Connections: USB 3.1 (gen 1) port and cable (SuperSpeed, 5Gb/s transfer speed)
Included cable: USB Type-C to Type-A. Coral, a division of Google, empowers innovative AI ideas with its local AI platform.
accelerator | 4 TOPS (int8); 2 TOPS per watt |
Connector | USB 3.0 Type-C* (data/power) |
Dimensions | 65 mm x 30 mm |
Host system support | Linux, Mac and Windows |
Peak power consumption | up to 900 mA @ 5 V |
USB | USB 3.0 (USB 3.1 Gen 1) Type C receptacle |
Standard USB disk size | 65 mm x 30 mm x 8 mm |
Operating frequency | 25°C - Reduced; 25°C - maximum |