UNIHIKER K10 low-cost TinyML education platform supports image detection and voice recognition

UNIHIKER K10 is a low-cost STEM education platform for TinyML applications that leverages the ESP32-S3 wireless microcontroller with vector extensions for workloads such as image detection or voice recognition.

The UNIHIKER K10 also features a built-in 2.8-inch color display, a camera, a speaker, a 2-microphone array, a few sensors, a microSD card, and a BBC Micro:bit-like edge connector for power signals and GPIOs. It’s a cost-optimized version of its Linux-based big brother – the UNIHIKER M10 – first unveiled in 2022. Arnon also reviewed the UNIHIKER in 2023, showing how to configure it, use the SIoT platform with MQTT message, and program it with Jupyter Notebook, Python, or Visual Studio Code. Let’s have a closer look at the new ESP32-S3 variant.

UNIHIKER K10 specifications:

• Core module – ESP32-S3-WROOM-1
  • MCU – ESP32-S3N16R8 dual-core Tensilica LX7 up to 240 MHz with 512KB SRAM, 8MB PSRAM, 16MB flash
  • Wireless – WiFi 4 and Bluetooth LE 5
  • PCB antenna
• Storage – MicroSD card slot
• Display – 2.8-inch 240×320 color LCD (ILI9341 driver)
• Camera – 2MP camera with 80°FOV (GC2145 sensor)
• Audio
  • 2W speaker
  • 2x MEMS microphones
• USB – 1x USB-C port for power and programming
• Sensors
  • AHT20 temperature (-40-85°C) and humidity (0 to 100%RH) sensor
  • LTR303ALS ambient light sensor (up to 64k Lux)
  • SC7A20H accelerometer
  • 2x MEMS MIC
• Expansion
  • 3-pin Gravity connector for 2x GPIO
  • 4-pin Gravity I2C connector
  • Edge connector
    • 3x GPIO (for banana clips)
    • 15x Digital IO
    • I2C
    • 3.3V and GND
• Misc
  • A, B, RST, and BOOT buttons
  • 3x WS2812 RGB LEDs
• Power Supply
  • 5V via USB-C port
  • 3.3V via edge connector
  • 2-pin connector for battery
• Dimensions – 83 x 51.6 x 11mm

The UNIHIKER K10 comes with pre-installed AI vision and audio models: face detection, image recognition, cat/dog detection, QR code recognition, motion detection, local (e.g. offline) speech recognition, and custom voice commands. You can switch between the different modes using the A/B buttons on the model.

Demos are nice, but it would not exactly be an education platform, if students can’t learn to program. This can be done in two ways: either with the MindPlus visual programming IDE available for Windows, macOS, and Linux,  and with a MicroPython firmware plus an IDE like Thonny. You’ll find all resources to get started and code samples on the wiki.

DFRobot says the UNIHIKER K10 is suitable for grade 3 or above (8+ years old), while the more powerful UNIHIKER M10 is better suited for grade 6 or above (11+ years old) as students can also used more advanced models like Yolo, MobileNet, MediaPipe, etc…

The UNIHIKER K10 can be purchased on DFRobot for $28.90 plus shipping. This compares to $79.90 for the original UNIHIKER (M10), although you can get a bundle with both the K10 and M10 for $79.90 until April 15 as part of a time-limited promotion.

Thanks to TLS for the tip.

Support CNX Software! Donate via cryptocurrencies, become a Patron on Patreon, or purchase goods on Amazon or Aliexpress