"These facial recognition and analysis tools constitute the entry point into a myriad of smart services within the connected home"
From research to management… and back. After having managed one of Orange’s research areas around customer experience, Stéphane Pateux made the bold decision at the end of 2015 to return to research as a contributor, in the area of artificial intelligence.
“My managerial experience was rich and interesting, but I really enjoy the technical side. I wanted to carry on taking part in the technological development of the Group, and to bring on board my research experience“, he tells us.
An engineer from the Corps des Télécommunications, Stéphane Pateux joined Orange in 2004 after spending nine years at the INRIA (the French Institute for Research in Computer Science and Automation) as a research fellow for video and network video transport analysis techniques. He took part in the development of video compression standards, including the recent HEVC coding standard, which has enabled a considerable reduction in video streaming transmission rates and the introduction of ultra-high definition into networks.
Deep learning
Now Stéphane Pateux is working on artificial neural network technologies, which have evolved considerably over the past few years thanks to the addition of more and more layers of neurons, thus contributing to the rise of deep learning.
Reproducing the operation of biological neural networks in a simplified manner, these technologies aim to teach computers to see and to hear – and to understand what they do see and hear! – through experience.
With his team, Stéphane Pateux is developing, for example, facial recognition and analysis tools that enable faces to be found in an image or a video, to know a person’s gender, to estimate their age, to identify, them, and even assess their emotions.
Presented at the Salon de la recherche 2017, his work fits into Orange’s vision of a future “sensitive home” and provides content for the Group’s offer, in particular around a home assistant.
“It’s the tools that enable you to be immediately recognised by your home; the entry point, in a way, to a myriad of smart services within the connected home” Stéphane Pateux explains.
In practice, facial recognition and analysis techniques enable the personalisation and adaptation of these services to each occupant. If we take the example of video on demand (VOD), a child will not be offered the same catalogue of films as an adult.
Publications and awards
These works and these solutions have been the subject of several publications and awards. “We are integrated in the research ecosystem, and are recognised through the quality of our work, Stéphane Pateux underlines. For example, our technology on gender and age estimation was rewarded during an international competition last year. It was judged higher performing than a human being.”
If contact with the academic world is frequent and fruitful, so is that within Orange, where Stéphane Pateux’s team interacts in particular with the “Anticipation” teams of the various Orange Labs.
Having known both the academic research world and that of industrial research, the researcher underlines the contributions of the latter, in particular at Orange: “Within the scope of industrial research, we have to make technological choices based on a techno-economic approach” he explains.
“Thanks to the quality of the different researchers and engineers working here, we are able to analyse the solutions and to qualify their performance in order to make these choices, which must be viable for the Group whilst meeting its customers’ expectations. We do not push technologies for the pure beauty of technology; we truly try to find out how they can serve users and if it is the right time to introduce them.“
