New 1.5” AI Vision Board Runs Years on a Single Battery

Eta Compute has unveiled its latest embedded AI vision board, the ECM3532, promising years‑long operation on a single battery thanks to its ultra‑low‑power design.

With a footprint of just 1.5\" × 1.5\", the ECM3532 packs an integrated battery, IoT and Bluetooth Low‑Energy radios, making it ideal for prototyping, field trials and production deployments of vision‑enabled devices. The board ships with an ambient‑light sensor, a microphone, an accelerometer/gyroscope, the low‑power Himax HM0360 camera, and an expansion connector.

The board’s energy‑efficient operation removes the constraints of tethered setups and short‑lived, high‑power boards, enabling continuous, autonomous vision workloads.

This release follows the first generation of Eta Compute’s modular AI solutions, underscoring the company’s commitment to low‑power, edge‑centric design.

Edge Impulse’s ML platform powers rapid neural‑network development on the board, while the new TENSAI Flow software from Eta Compute streamlines the entire workflow—from concept to firmware—by providing a neural‑network compiler, a curated zoo, and full FreeRTOS‑based middleware that simplifies sensor and cloud integration.

Benchmarking on the same CIFAR‑10 workload shows the TENSAI compiler reduces energy per inference by 54× compared to a direct implementation on a competing SoC, and by 200× when paired with the TENSAI zoo model, illustrating the platform’s efficiency gains.

TENSAI Flow’s secure data pipeline with Edge Impulse lets developers train once and deploy real‑world models for future refinement. The compiler automatically optimizes TensorFlow Lite models for the TENSAI SoC, delivering peak power efficiency while supporting seamless sensor integration. The tool also provisions cloud connectivity and OTA firmware updates based on new models or data.

Jeff Bier, founder of the Edge AI and Vision Alliance, praised the breakthrough: “Computer vision turns devices into perceptive agents, yet heavy computation has limited its reach. Eta Compute’s innovation—paired with other Alliance members—makes vision practical for power‑constrained scenarios.”

Zach Shelby, CEO of Edge Impulse, added, “With our joint platform, developers can swiftly prototype and deploy a spectrum of vision workloads—from object detection to counting—across humans, animals and machines.”

Ted Tewksbury, CEO of Eta Compute, concluded, “The integrated board, coupled with Edge Impulse’s development environment, empowers teams to launch vision solutions that can profoundly impact people’s lives and businesses.”