Eta Compute's ECM3532 AI Vision Board Promises Even Tinier ML
ECM3532 NSP with on-board camera and sensors brings ultra-low-power machine vision to the edge.
The world of machine learning keeps getting tinier and tinier — and with their second ECM3532-based board, Eta Compute has shrunken machine vision applications to a diminutive 1.5 square inches! Based on the same silicon as their AI Sensor Board, the new AI Vision Board features an on-board Himax HM0360 (as featured on Himax's own WE-I Plus EVB Endpoint AI Development Board), as well as tight integration with Edge Impulse Studio, enabling rapid development of low-power AI vision solutions.
In addition to the Cortex-M3-based ECM3532 Neural Sensor Processor, the AI Vision Board is equipped with an ambient light sensor, microphone, and accelerometer/gyroscope — as well as an expansion connector for power and UART communication. In case you don't already have one lying around, Eta Compute thoughtfully include a SparkFun FTDI Basic Breakout (3.3V! The AI Vision Board is not 5V tolerant!), since there is no on-board microUSB or other dongle-free programming interface (advanced users can also take advantage of the integrated SWD CoreSight connectors for the ECM3532 and A31R118 BLE). Bluetooth Low Energy facilitates low-power communication with the device, and a microSD card slot provides additional RF options, in addition to storage. And speaking of storage, the board features a whopping 64 Mbit of serial flash, meaning plenty of space for your largest tinyML models, or extensive data logging! A CR2032 coin cell battery holder, red and blue LEDs, and an action button round out the package, giving you everything you need for a wide range of applications, on a single, power-optimized board.
Despite its small size and low power consumption, the AI Vision Board is capable of some mighty tasks. Applications include sound classification and keyword spotting using the built-in microphone, gesture and defect detection using the accelerometer/gyroscope, and a slew of vision-based tasks such as image classification as well as person and object detection and counting.
A look at its key features:
- ECM3532 Neural Sensor Processor with hybrid multi-core Cortex-M and DSP operating up to 100MHz with continuous voltage and frequency scaling technology (CVFS)
- 512kB embedded Flash
- 256kB + 96KB SRAM
- Himax HM0360 camera
- Ti OPT3001 ambient light sensor
- TDK ICS-41350 MEMS microphone
- TDK ICM-42688-P MEMS accelerometer/gyroscope
- ABOV A31R118 BLE v4.2 with antenna
- 64Mbit serial Flash
- CR2032 battery socket
- 6-pin UART and power port
- microSD extension for other types of RF
- 2 LEDs
- 1 push button
- < 1mw inference for machine learning algorithms
Get to know the Eta Compute ECM3532 AI Vision Board at Hackster's Ultra Low Power Computer Vision Applications with Edge Impulse and the Eta Compute AI Vision Board webinar on March 31st, 2021 at 12pm ET/9am PT, and be sure to pick one up before the end of March at an introductory 50% off discount!