It also features an IMG PowerVR GE8300 GPU that's integrated in the SoC, while the machine-learning accelerator consists of the Google Edge TPU coprocessor that's capable of performing four trillion operations per second (TOPS) or two TOPS per watt.
According to Google, it can execute mobile vision models such as Mobile:Net v2 at almost 400 frames per second. The device runs a derivative of Debian Linux called Mendel.
The computer also comes with 2GB LPDDR3 RAM and 8GB of eMMC flash memory. Wireless connectivity includes Wi-Fi 5 802.11a/b/g/n/ac and Bluetooth 5.0.
There's a 3.5mm audio jack, digital PDM microphone, 2.54mm two-pin speaker terminal, while video output includes a micro HDMI 1.4 port.
For the camera, there's a 24-pin four-lane FFC connector for MIPI-CSI2 camera and a 39-pin FFC connector for a four-lane MIPI-DSI display.
Additionally, there's a 40-pin GPIO header and two USB Type-C (USB 2.0) ports.
The Dev Board mini supports the TensorFlow Lite machine-learning framework and can be coded in Python and C++. It also works with Google's AutoML Vision Edge. Google views it as an evaluation device for the Accelerator Module – a surface-mounted module that provides the Edge TPU. However, it's also an embedded system that can be used for on-device ML projects.