CONTEXT
Image segmentation is necessary for many applications and in many fields: medical imaging, quality analysis, autonomous car ... The classic techniques used today are diverse and varied and have two major drawbacks: their precision and their time calculation.
DESCRIPTION
Deep learning techniques provide a relevant solution to these drawbacks. The automatic segmentation of images by deep learning requires, as a first step, the design of a suitable neural network architecture. By using a large database of good quality and labeled images, it is possible to train the neural network to learn to perform certain tasks automatically. Once this network has been trained, the results obtained are precise and the calculation times are short. What is more, this network can be re-trained at will, as soon as, for example, new data is available allowing an improvement over time in its performance.
COMPETITIVE ADVANTAGES
MARKETS AND APPLICATIONS
DEVELOPMENT STAGE
Technology validation in an operational environment (TRL 6)
RESEARCH TEAM
Institut FEMTO - Sciences & Technologies
INTELLECTUAL PROPERTY
Software deposit in progress
TARGET PARTNERSHIP
Software license or co-development
CONTACT
Abdelkader GUELLIL
Business Development Manager
+33 (0)6 26 61 89 06
abdelkader.guellil@sayens.fr