Matrox Imaging Industrial Research Chair in Computer Vision for Industrial Applications
About the Chair
Computer vision is a fast-growing aspect of artificial intelligence that will continue to play an increasingly larger role in many key sectors, including automated manufacturing. However, deep learning techniques must be significantly improved in order for this technology to progress from the lab to the industrial world. This is the mission undertaken by the Matrox Imaging Industrial Research Chair in Computer Vision for Industrial Applications.
The objective of our work is to resolve the issue of collecting and labelling massive volumes of data, to enhance the capacity of existing models to adapt to changing conditions within the application environment, to facilitate the interpretation of data presented in the form of images by deep learning systems and to integrate complex rules into the models to significantly improve the accuracy of current approaches based on deep learning.
More specifically, our program will lead to the following:
- The design of specialized computer vision algorithms that are capable of learning despite a limited quantity of annotated data, while quickly adapting to fluctuations in the application environment;
- The creation of new object recognition models that are capable of processing both 2D images and 3D point clouds at the same time, improving their performance by combining the two types of data;
- The development of an optical character recognition (OCR) system that is capable of dynamically integrating rules into the models to be recognized and adapting to character fonts that have not yet been observed during the learning process.