AND

AND TEAM: Distributed digital algorithmics

Context

The research activities of the AND team are articulated around 3 main axes and a transversal one. The main axes are: image processing on specific architectures, security and encryption and distributed algorithms for IoT. The transversal axis deals with artificial intelligence and deep neural networks. This cross-disciplinary approach is characterized by several works in which the contribution of the latest Artificial Intelligence techniques has led to significant advances (image processing, watermarking, data compression for the IoT, etc.).

Goals and Research Areas

The AND team’s research is concentrated on the 4 following themes:

› Artificial intelligence and deep neural networks

This transversal axis focuses on different fields of application:

•  Algorithms for objects classification and image processing. We have designed image de-noising algorithms, and different technics for restoring compressed images or improving their quality.
•  Algorithms for natural language processing, for example the analysis of hospital patient records for automatic classification.
• Algorithm for time series prediction, for example we are interested in the prediction of fire department interventions.
•  Algorithms for medicine and bioinformatics. With the Bourgogne CHU de we have developed algorithms for the segmentation of the myocardium and the different parts of the heart in case of infarction.

› Image processing on specific architectures

Objectives : In order to benefit from the computing power offered by architectures such as GPUs and FPGAs, the team is working on designing and deploying high-performance image processing algorithms on these architectures. In this way, we have designed algorithms for the calculation of median filters (separable or not), convolution filters, etc.

› Security and cryptography : The team's work concerns the design of robust encryption algorithms or hashing technics adapted to computing architectures ranging from the IoT to the GPU. Other work concerns steganography/steganalysis or watermarking techniques

› Distributed algorithms for the IoT : One of the major issues of the IoT is the conservation of sensor energy. The researchers of the team are developing efficient distributed algorithms for the acquisition and processing of data by sensors. The works are also focused on minimizing (very energy-intensive) communications by performing edge preprocessing before performing larger calculations in the cloud.

Implementing our work

The team works with many partners: Hôpital Nord Franche-Comté, Service Départemental d'Incendie et de Secours du Doubs (SDIS 25), Colruyt, Faurecia, Wudo, Caduciel, ...