tinyML Talks on July 20, 2021 “Better productivity leveraging AI community driven interoperability” by Danilo Pietro Pau

We held our next tinyML Talks webcast. Danilo Pau from STMicroelectronics presented Better productivity leveraging AI community driven interoperability on July 20, 2021.

July 20 forum

Is artificial intelligence a technological fad? Why are resource-constrained AI and tinyML important? Reviewing the milestones of AI and learning more about the benefits of resource-constrained AI is essential to overcome the limitations of a centralized approach to AI. To address the challenges and opportunities in data processing closer to sensors and in real time, ST offers a unique 5-step method with a set of related tools to automatically deploy pre-trained neural networks on IoT (STM32) and Automotive (SPC58) microcontrollers, as well as machine learning for in-sensor computing.

One year before graduating from the Polytechnic University of Milan in 1992, Danilo PAU joined STMicroelectronics, where he worked on HDMAC and MPEG2 video memory reduction, video coding, embedded graphics, and computer vision. Today, his work focuses on developing solutions for deep learning tools and applications. Since 2019 Danilo has been an IEEE Fellow. Currently serves as a member of IEEE Region 8 Action for Industry and Member of the Machine Learning, Deep Learning and AI in the CE (MDA) Technical Stream Committee IEEE Consumer Electronics Society (CESoc). With over 80 patents, 104 publications, 113 MPEG authored documents and 39 invited talks/seminars at various worldwide Universities and Conferences, Danilo’s favorite activity remains mentoring undergraduate students, MSc engineers and PhD students from various universities.


Watch on YouTube:
Danilo Pau

Download presentation slides:
Danilo Pau

Feel free to ask your questions on this thread and keep the conversation going!

Using the Studio… Do we need a license?
No, tools are free of charge. Download them @

What is a good first step(s) for someone with an understanding of the embedded world who wants to get into tinyML?Start with edX courses on TinyML

Does ST plan to continue support of TinyML developer tools outside of Cartesium?
With X-CUBE-AI yes

I am new to TinyML so I ask, how to understand the career prospects of TinyML and are there any communities to meet people interested in TinyML? www.tinyML.org, also tinyML Community on LinkedIn

What are the cons and pros of having extra resources (i.e. External RAM) to the stm32 devices in order to use complex models ?
Using external RAM implies more power consumption and higher latency. Therefore it is a precious resource to be used. However any possible effort has to be done to reduce model complexity and quantize the model to save resources maintaining the accuracy at the expected level

Is there a simple way to compare different NNs for the given task to produce best mix of memory, latency and power consumption?
This is the focus of TinyMLPerf, which standardize benchmarks to measure efficiency of NN eimplementations on low power systems

Sure things… Thanks alot
Appreciate your time and the one of all attendees in listening my talk.
Danilo Pietro Pau