Thesis in cooperation with Télécom ParisTech - Cifre thesis

DxOMark is the most trusted reference in photo and video quality for the mobile industry and most known camera manufacturers.

Our experts at DxOMark provide consulting services at all stages of development of imaging systems, from the initial specifications to the fine- tuning post-treatment algorithms.

We also lend our world-renowned technical expertise to the normative efforts of international standard-setting bodies such as the ISO and the IEEE, as well as to publications in international journals and conferences.

Image quality assessment is one of the core skills of our company and we have developed a large expertise on this topic. In order to assist our experts and eventually automatize some tasks, we consider Machine Learning and in particular Deep Neural Network (DNN) as natural and promising track to explore.

Deep Neural Networks have in recent years been the subject of great attention as solutions for many complex problems previously requiring direct human intervention by experts.

In one of the most successful examples of deployment, top-end smartphone making companies have been integrating more and more DNN techniques to enhance the aesthetic qualities of the photos and videos taken with their devices.

While opening a range of new exciting possibilities in mobile photography, this adds more technical facets to considerate for those actors in the industry that provide imaging systems manufacturers with automated solutions to assess and quantify image quality. This is because a number of long-held assumptions are no longer true, namely that the behavior of the device depends only loosely on the content of the scene and not at all on its semantics, thus allowing a de-correlated evaluation of its different image quality attributes.

DxOMark Image Labs is uniquely well poised to make great strides in this field thanks to our unparalleled database of images and video, labelled according to a wide range of image quality attributes, as well as our decades-long experience in the field.

The candidate shall be integrated part-time in of one of our inclusive agile teams of image quality experts, developers, and technicians. Direct cooperation with our talents is encouraged.

The candidate will also be required to periodically deliver proofs-of-concepts and prototypes and will have the opportunity to work together with our operation teams to field test her/his tools against real-word data coming from the cutting-edge technology of our customers.

Profile : DxOMark is looking for students with a Master or Engineering degree in machine learning or image signal processing that have firm theoretical understanding as well as good programming skills, ideally with an experience in deep learning.

English fluency is required as well as French communication capability.