The leader in energy-efficient endpoint AI solutions, Eta Compute, has announced the ECM3532 AI Vision board, the most energy-efficient vision board, to speed up machine learning operations. The company’s second AI board is based on the ECM3532 TENSAI SoC with a Cortex-M3 microcontroller and a separate CoolFlux DSP. This new AI Vision board features an onboard Himax HM0360 camera. It comes with tight integration with Edge Impulse Studio, enabling the rapid development of low-power AI vision solutions.
Many boards in the rising world of AI-enabled development boards, such as the Arducam Pico4ML Board or the high-performance BeagleV Development Board. Compared to these boards, the ECM3532 AI Vision board offers a pretty small form factor of just 1.5 square inches and a shallow power usage of around 1mW for ML applications. This gives it an edge over other boards in applications where power consumption is critical.
“Our unique integrated ECM3532 AI Vision Board simplifies and speeds development time while reducing risk and costs for developers creating AI vision solutions,” said Ted Tewksbury, CEO of Eta Compute. He explains that “For the first time, they can rely on an integrated board complemented by Edge Impulse’s machine learning development platform to deploy vision applications that can have the power to transform people’s lives and work.”
A look into ECM3532 AI Vision Board’s features and form factor
The board features a Cortex-M3-based ECM3532 Neural Sensor Processor as its main MCU. It can operate with frequency up to 100MHz with self-timed continuous voltage and frequency scaling technology (CVFS). Along with that, it comes with 512 kB of embedded Flash and 352 kB of SRAM. It also features a CoolFlux DSP16 which is an ultra low power 16-bit DSP.
As for the onboard sensors, the main highlight is its vision sensor which is a Himax HM0360 camera. Other than that, there’s one PDM MEMS Microphone, one 6-axis MEMS Accelerometer/Gyroscope, and a Texas Instruments OPT3001 Ambient Light Sensor. The board also provides numerous peripherals, including UART, SPI, I2C, I2S, PDM, ADC, and Timers.
Block diagram of ECM3532 AI Vision Board
Coming to the AI Vision board’s hardware features, there’s a battery cradle for a CR2032 battery. It supports Bluetooth Low Energy on board. It also has an extension for different types of RF through a Micro SD card slot. There are also two LEDs (Red and Blue) and a push-button. They are very useful for debugging or system monitoring purposes.
Talking about the software part, the board can easily be programmed using Edge Impulse’s machine learning development platform. This platform provides an excellent GUI for building fast neural network development, making energy-efficient vision endpoints seamless. Eta Compute also includes a SparkFun FTDI Basic Breakout -3.3V board in the box, which allows for programming through micro-USB.
This board’s applications mainly revolve around the embedded vision: image classification, person detection, object, and people counting. The board is also capable of running sound and motion algorithms like sound classification, keyword spotting, activity classification, context awareness, and defect detection.
The AI Vision board retails for $130 on its official product page. All the images used are also taken from the same page.
Harsh Chaudhary is an engineering student currently pursuing Electrical Engineering. He’s a robotics and tech enthusiast and likes to write about stuff related to IoT and embedded systems. His vision is to use Robotics to make the life of humans easier.
6 Responses
This is a great board, have you guys tried it. The amazing thing how much less power it consumes 13uA per Mhz. Ambiq micro has very low power MCU’s as well(3uA/Mhz).
Yes, the features which it offers in such a small size are really impressive. We haven’t tried the board yet.
Do you know if the controller is available from the company or from their distributors? Any Idea about the cost?
By controller, you mean the ECM3532 vision board right? If yes, then its available for $130 and the buying options are mentioned on their product page (https://etacompute.com/tensai-ai-vision-board/)
I mean the MCU itself, we can evaluate with the board but then, for our application we would want to make our own custom board, right?
130$ is not reasonable for production.