Eta Compute's Tensai Flow Puts Machine Learning at the Edge of the IoT
August 11, 2020
Deploying artificial intelligence and machine learning at the Edge of the IoT has long been the Holy Grail for design engineers.
Deploying artificial intelligence and machine learning at the Edge of the IoT has long been the Holy Grail for design engineers. In most cases, there simply wasn’t enough compute power to tackle such complex operations, often with limited power resources available. Thanks for tools like Tensai Flow, the software suite in Eta Compute’s Tensai Platform, developers can now implement such systems.
Tensai Flow, which enables seamless design from concept to firmware, includes a compiler, a neural network zoo, and middleware with FreeRTOS, a hardware abstraction layer (HAL) and frameworks for sensors and IoT/cloud enablement.
The software suite complements the company’s existing resources to speed applications development. According to the company, the software addresses all aspects of designing and building a machine learning application for IoT and low power edge devices. This includes a reduced memory footprint, fewer operations, and overall less complexity.
A key feature of the Tensai software is its ability to reduce development risk by confirming feasibility and proof of concept. The neural network zoo accelerates and simplifies development with ready-to-use networks for the most common use cases, including motion, image, and sound classification; developers simply train the networks with their data.
One result is that TensorFlow networks can run on Eta Compute's ultra low power SoC. Testing has shown AI performance in the 1-mW range, which is quite low compared to other alternatives, particularly those designed for image processing.