Machine Learning and Deep Learning are often presented as technologies reserved for large enterprises and data centers. But did you know that this kind of technology can be deployed on a small scale? We have experienced this a few times ourselves. For example, we used a modest model with only 150 parameters to analyze the topology of an object using a few imprecise distance sensors, and provide a value characterizing this topology. We trained the model with a few thousand measurements over a few hours, and finally our results were impressive, with almost a decade of additional precision compared to an analytical approach. And all of this does not require servers or data centers - a small microcontroller costing only a few euros is enough for all the steps, from training to use, without any Internet connection.

If you too are interested in these kinds of solutions for your business, contact us!