The Nanoelec/Smart Imager program is committed to providing relevant and real-time environment analysis by vision. It could be one of the greatest technical challenges at Nanoelec for the coming years. Overview by Eric Ollier, Program Director.
What are the opportunities for smart imagers?
There is a growing number of applications where sensors become a vehicle for decision and action. There are market opportunities for consumers, smart buildings, healthcare, industrial process flows and monitoring, automotive and robotics. Mastering smart imagers industrial process flow would be a competitive advantage for Nanoelec’s industrial partners.
Tell us why this program represents a real breakthrough and paradigm shift?
For the development of smart imagers, computing and data storage are the main challenge, in addition to image collection. Several technologies –CMOS image sensors, 3D stacking, AI, Neural Networks– are now mature enough to address new possibilities. A wide range of processing functions is possible, including multi-spectral management and associated data fusion. We have to ‘wrap-up’ these possibilities to build new functions in multispectral image analysis for example. To do that we must develop and combine the best 3D stacking technologies and also imagine the best architecture and partitioning between functions, including neural networks embedded in the smart sensor.
How does artificial intelligence integrate imagers?
To make the leap from image generation to information extraction, we intend to embed artificial intelligence processing inside the sensor. And this will require a dedicated embedded computing die. So we have to think about a new hardware and software architecture capable of enabling the implementation of a neural network and additional digital processing inside the smart sensor.
What is your strategy?
Our main goal is to evaluate the benefits of 3D-stacking technologies. This encompasses a wide range of innovations in silicon technologies – process and characterization, design, design tools, architectures, software engineering – as well as in the field of software.
What in the Nanoelec background enables it to start such a challenging new program?
The previous “3D integration” program has been very active in terms of technological developments, publications and IP. The partners of the Smart Imager program intend to use this legacy and pursue this policy actively, especially to strengthen the foundations of our stacking technology for new 3D architectures.
Form your personal point of view, what is the most exciting part of the program?
I think what is most exciting is the fact that we have to find practical and industrial solutions based on the most advanced technologies to embed artificial intelligence in the smart imaging sensor. It is also highly motivating to work in very close collaboration between R&D and industrial partners, working with various teams involved in hardware technologies but also in design, design tools, architecture, AI… Finally, it’s very exciting to prepare the next generation of smart imaging sensors!