Eta Compute has launched an integrated artificial intelligence (AI) embedded vision board which the company claims can enable vision applications that can last for years on a single battery.
Its ECM3532 AI vision board’s small form factor (1.5”x1.5”), embedded battery and low-power internet of things (IoT) and Bluetooth Low-Energy connectivity make it suitable for prototyping, field testing, and deployment of AI embedded vision applications. The board includes three sensors (ambient light, microphone, accelerometer/gyroscope), a low power Himax HM0360 camera, and an expansion connector.
The company said its ultra-low-power operation eliminates barriers with traditional, tethered solutions or boards that have extremely limited battery life and high-power consumption. The AI vision board is the second in its family of boards, modules, and systems designed by Eta Compute.
The new board is supported by Edge Impulse’s machine learning (ML) development platform for fast neural network development making the design of energy-efficient vision end-points seamless. The companies collaborated to integrate Eta Compute’s TENSAI Flow software, optimizing the design flow for efficiency in embedded AI design of next generation intelligent devices. This software allows developers to quickly verify feasibility and proof of concept, and enables seamless design from concept to firmware, for creation of ML applications in IoT and low power edge devices. TENSAI Flow includes a neural network compiler, a neural network zoo, and middleware comprising FreeRTOS, hardware abstraction level (HAL) and frameworks for sensors, as well as IoT/cloud enablement.
Compared to direct implementation on a competitive device of the same CIFAR10 neural network, the TENSAI neural network compiler on TENSAI SoC (system on chip) improves energy per inference by a factor 54x. Using the CIFAR10 neural network from TENSAI neural network zoo and TENSAI neural network compiler improves the energy per inference further, bringing this figure to a 200x factor.
Through its interface with Edge Impulse, TENSAI Flow allows developers to securely acquire and store training data so customers train once and have real-world models for future development. The software automatically optimizes TensorFlow Lite AI models for Eta Compute’s TENSAI SoC, delivering high optimization and power efficiency. With TENSAI Flow, TENSAI SoC can load AI models that include sensor interfaces seamlessly. TENSAI Flow provides the foundation to automatically provision and connect devices to the cloud and upgrade firmware over the air based on new models or data.
Explaining the significance of low power embedded vision capability, Jeff Bier, founder of the Edge AI and Vision Alliance, said, “With computer vision, devices can understand the world around them, enabling them to be more capable, safer, easier to use and more autonomous. But vision algorithms are very compute-intensive, and the power consumption required to deliver the necessary processing performance has made these capabilities impractical for many potential applications. We applaud Eta Compute’s innovation and collaboration with other Edge AI and Vision Alliance companies, which is making vision feasible in many new power-sensitive applications.”
The CEO and co-founder of Edge Impulse, Zach Shelby, commented, “Together with Edge Impulse, Eta Compute developers can design, test, and deploy rapid embedded applications across a multitude of workloads from object detection to classification, and to actual counting, across the human, animals, and machine spectrums.” Ted Tewksbury, CEO of Eta Compute, added, “For the first time, they can rely on an integrated board complemented by Edge Impulse’s machine learning development platform to deploy vision ap-plications that can have the power to transform people’s lives and work.”
- Key technologies power growing role for embedded vision
- AI vision processor enables 8K video at 30fps in under 2W
- Smart camera offers turnkey edge machine vision edge AI
- Embedded Vision Summit: ML perception on tiny FPGAs
- CoaXPress 2.0 chips enable machine vision at 12.5Gbps on single cable
- Stacked sensor architecture brings advanced vision capabilities