- ABOUT IMC
- IoT LIBRARY
Ready Models provide swift way to ML production
- December 6, 2023
- Steve Rogerson
Swedish edge AI company Imagimob has introduced Ready Models, which it says is the fastest way to take machine learning to production.
The Ready Models let companies add edge AI features to their smart devices with a fraction of the resources. They can be quickly deployed onto existing microcontroller (MCU) hardware, such as Infineon’s PSoC 6, without the cost, time or expertise required for custom development.
“If you look at the edge AI space right now, you can probably count on one hand how many companies provide off-the-shelf models for any one option,” said Sam Al-Attiyah, head of customer success at Imagimob “Our Ready Models are built upon eight years of expertise and thoroughly tested out in the field in different environments, so they are validated in terms of performance. And the fact that we are running them on small edge devices is really unique.”
To ensure the robustness of the models, the company has created a comprehensive list of different scenarios they can encounter, and then tested those scenarios. It also tests the models in different scenarios all over the world to ensure they work with no bias based on specific geographies or ethnicities.
Finally, field testing on device lets it test and document realistic model performance in an expected hardware setup. The result of all this is a model that does exactly what the user wants it to when it is running in its products.
Imagimob has launched four audio-based Ready Models, including Baby Cry for baby monitors, Siren Detection for pedestrians, and Coughing Detection and Snoring Detection, which are both for wearable devices in the medtech and health sectors.
Additional models are under development within audio, radar, IMU and capacitive sensing.
For many companies seeking to upgrade products with smart AI features, up until now the barriers have been too high. The typical development process for a custom-made ML model not only requires the right software engineering and AI expertise, but also investment of time and resources for a development process that spans from collecting, validating and labelling data, to training models based on those data, deploying them on device, and then testing them in diverse environments to ensure performance expectations are met.
In contrast, Ready Models require zero or limited engineering and AI competence to implement. And with all the development and testing work already put in, they essentially offer a shortcut to the marketplace.
“This is a much easier way for companies to begin their ML journeys,” said Anders Hardebring, CEO of Imagimob. “They don’t have to make such a big investment to start using it on their edge devices. Depending on the skillset and expertise in a company, developing a custom model for production typically takes six months to a year. With our Ready Models, they can have new edge AI features up and running essentially overnight.”
Based in Stockholm, Sweden, Imagimob (imagimob.com) has been serving global customers within the automotive, manufacturing, healthcare and lifestyle industries since 2013. In 2020, it launched the Imagimob Studio platform for the swift and easy end-to-end development of edge AI applications for devices with constrained resources. Imagimob has been a part of Infineon Technologies (www.infineon.com) since May 2023.