Last April, we wrote about Edgeless EAI-Series dual Arm Cortex-M4 MCU equipped with a 300 GOPS CNN-NPU for AI at the very edge as we had discovered the chip in an upcoming Banana Pi board.
It turns out Banana Pi BPI-EAI80 development board powered by Edgeless EAI80 AI microcontroller has just launched for $16 on Aliexpress, or you could get a complete kit with a touchscreen display, a camera, and a USB power supply for $80.
- System-in-Package – Edgeless EIA80 dual-core Cortex-M4F microcontroller @ 200MHz with 300GOPS AI accelerator (CNN-NPU), 384KB of SRAM including 256KB for CNN-NPU, and 8MB SDRAM
- Storage – SPI flash
- Display I/F – LCD connector up to 1024×768
- Camera I/F – 1x DVP camera interface
- Audio – 2x onboard microphones
- Connectivity – 2.4GHz 802.11b/g/n WiFI 4 using ESP8266 module
- USB – 1x USB 2.0 Type-C port
- 40-pin GPIO header (multiplexed with LCD interface)
- CANBUS 2.0 A/B header
- Debugging / Programming – Debug UART header, SWD header, and ESP8266 programming header
- Misc – Reset and wakeup buttons
- Power Supply
- 5V via USB-C port
- 3.7V Li-ion battery via 2-pin header
- Dimensions – 86x54mm or about the same as Raspberry Pi Model B board
The complete development kit includes a BPI-EAI 80 board, a 7-inch touch screen display with 800×600 resolution, a 2MP OV5640 camera module, and a 5V/2A USB Type-C power supply.
The board runs uGelis embedded OS and can be programmed using the Kelis SDK that includes AI audio and computer vision demos for word recognition, hand gesture detection and recognition, and human body detection. Hardware and software documentation can be found in several PDF files on the Wiki, but I could not find any easy getting started guide, so users may require some studying before having a demo setup and running. Edgeless is a subsidiary of Gree Electric Appliances, that happens to be the world’s biggest air-conditioner manufacturer, so support might be better than for smaller companies.
Edgeless EIA80 processor and Banana Pi BPI-EIA80 board are designed for voice control. completer vision, AIoT, home & building automation, motor control, industrial computing, and STEAM education.
The features and performance look similar to what you’d get with Kendryte K210 RISC-V AI processor equipped with a 230 GOPS AI accelerator and found in low-cost devkit such as Maixduino or TTGO watch.