Coral Dev Board Mini - single board computer with SoC MediaTek 8167s

Description

The Coral Dev Board Mini is a single board computer that provides high-speed machine learning (ML) inference with small board dimensions. It is primarily designed as an evaluation device for an accelerator module (the surface mount module that Edge TPU provides), but is also a full featured embedded system that can be used in a variety of ML projects.

The embedded Edge TPU coprocessor is capable of 4 trillion operations (tera-operations) per second (TOPS), consuming 0.5 watts for each TOPS (2 TOPS per watt). It can realize the most modern mobile vision models such as MobileNet v2, at almost 400 fps, with low power consumption. The computer has built-in WiFi 5 and Bluetooth 5.0 wireless connectivity, and numerous connectors for audio, video, camera, display and a 40-pin connector with numerous interfaces. The kit supports TensorFlow Lite models, so the user does not have to build ML models from scratch, and AutoML Vision Edge, which allows you to create accurate classification models for non-standard images. For the kit, the manufacturer provides full documentation with a quick start guide and numerous examples.

Features

• SoC MediaTek MT8167s chip:
  • Quad-core ARM Cortex-A35 processor:
    • Armv8-A 64-bit architecture
    • 32 KB of L1 instruction cache
    • 32 KB L1 data cache
    • 512KB Unified L2 Cache
    • Frequency up to 1.5 GHz
  • Imagination PowerVR GE8300 graphics processing unit (GPU):
    • 1.6 gigapixels / s at 600 MHz
    • It supports OpenCL 1.2 and Vulkan 1.0
    • Supports OpenGL ES 1.x, 2.0, 3.0, 3.2 3D
  • Video Processing Unit (VPU):
    • 1080p/60fps HEVC decoder
    • 1080p/60fps MPEG4 decoder
    • 720p/30fps MPEG-1, 2, H.263, H264 decoder
  • Built-in memory:
    • Boot ROM (96 KB)
    • Built-in RAM (256 KB + 64 KB)
  • Audio:
    • Interfaces support DAI, I2S, TDM, SPDIF
    • Supports 16-bit, 24-bit, mono, stereo formats
    • IIR 4-band compensation filter for improved speaker characteristics
  • Display:
    • HDMI 1.4a display output with resolution up to 1920x1080
    • MIPI-DSI display output 4 lane 24 pin DSI connector, maximum speed 1 Gb / s
  • Camera:
    • MIPI-CSI2 camera input
    • 4 lane
    • 24-pin CSI connector
    • Supports a camera up to 8MP
  • Security:
    • Built-in cryptochip (A71CH)
    • ARM TrustZone
• ML accelerator (Edge TPU coprocessor):
  • Efficient ML inference for TensorFlow Lite models
  • 4 trillion operations per second (TOPS)
  • 2 TOPS per watt
  • It uses a USB 2.0 interface with the SoC MT8167s
• Memory:
  • SDRAM memory:
    • 2 GB LPDDR3 SDRAM (4-channel, 32-bit bus width)
    • The maximum frequency of DDR 1600 MHz
    • Interfaces directly to the built-in DDR SoC controller
  • Flash memory (eMMC):
    • 8 GB NAND eMMC flash memory
    • 8-bit MMC mode
    • Compatible with JEDEC version 5.0 and 5.1
  • MicroSD slot:
    • Compliant with SD/SDIO 3.0 standard
    • Supports SD/SDHC/MMC cards
    • The maximum capacity of the card is 32 GB
    • Works in 4-bit SDIO mode
• Wireless network:
  • Wi-Fi 5 (802.11a/b/g/n/ac 2.4/5 GHz), supports SDIO3.0 interface for W-LAN
  • Bluetooth 5.0 (supports BLE), UART interface
• Support for peripheral devices:
  • 40 pin I/O connector
  • USB 2.0 Type-C OTG
  • USB Type-C powered
  • 2-pin mono speaker connector
  • 3.5mm headphone jack
  • Micro HDMI
  • MicroSD card
  • MIPI DSI Display (24 Pin Flat Flexible Cable)
  • MIPI CSI Camera (24 Pin Flat Flexible Cable)
• Dimensions: 64 x 48 mm