- M.2 accelerator module with Edge TPU designed to add local machine learning inference - Capable of performing 4 trillion operations per second (4 TOPS) with only 0.5 watts per TOPS - Enables efficient execution of advanced mobile vision models like MobileNet v2 at 400 FPS - Compatible with Debian-based Linux systems and M.2 B+M Key slots - Supports TensorFlow Lite models and AutoML Vision Edge for custom image classification - Easy integration into ARM, x86, mini-PC, and industrial systems - Ideal for AI-enabled NVR systems, home automation, and real-time object detection - Local processing reduces latency, improves privacy, and eliminates the need for constant connectivity