Edge Impulse provides a set of tools for developing machine learning applications to run on an edge device such as a microcontroller or FPGA. The tools take care of everything from creating a training dataset to model optimization.
AI and machine learning are notoriously difficult to implement in embedded systems at the edge: specialised applications often only have access to small sets of sensor-derived training data; the compute power available in embedded devices such as microcontrollers is relatively small in AI terms; and machine learning algorithms can require endless tweaking to produce reliable and accurate inferences.
Developers are in urgent need of technical solutions that will shorten the time between project start and the production of an effective machine learning system: this is why interest is growing in the Edge Impulse toolchain for machine learning at the edge. The advantage of Edge Impulse’s product is that it provides a comprehensive and integrated set of tools, resources and software for the entire development process.
The Edge Impulse tools start by guiding the collection of sensor data, for anything from image recognition – an Edge Impulse case study describes a project for building a portable machine for diagnosing diabetic retinopathy – to tracking assets in the cold chain. The Feature Engineering tool builds algorithms that pinpoint the features that matter in your application. Tuning tools then optimize the performance of the algorithm on your chosen hardware target. Edge Impulse speeds up application development by providing a broad range of pre-built algorithms developed for common machine learning use cases.
The tools are compatible with a huge range of SoC, microcontroller and microprocessor manufacturers, including some that ipXchange members will be familiar with: try using Edge Impulse to develop machine learning applications for your chosen Alif Semiconductor Ensemble microcontroller, an Arduino development platform, or the Syntiant neural processors.
Apply for the Syntiant board with built in Edge Impulse AI here:
Syntiant TinyML Development Platform for NDP101
Apply for Edge Impulse itself below:
You must be a member to leave a comment. Sign up or Sign in