IMDS Research Chair in Intelligent Multimodal Systems to Improve Productivity and Digital Document Security
About the Chair
The IMDS Research Chair in Intelligent Multimodal Systems to Improve Productivity and Digital Document Security is dedicated to developing AI-powered methods to automate document processing and to detect forgeries and tampering. Its goal is to help organizations process large volumes of documents more efficiently, reliably, and securely.
Invoices, IDs, payslips, and contracts are documents that combine text, images, and various layouts. The Chair is developing multimodal AI systems that can analyze these different sources of information en masse to extract data, classify documents, and identify signs of fraud.
Conducted in collaboration with IMDS (International Medias Data Services), the intent of the work is to design tools adapted to actual conditions of use, even in cases when document quality is variable, the documents use new formats, or there are few examples available for AI training.
Research areas
The research program is structured around four complementary areas.
Multimodal learning for document analysis
Develop AI models that jointly analyze textual content, layout, and visual elements, such as handwriting and stamps. This approach seeks to improve data retrieval, document classification, and detection of forgeries and tampering, even if some data is incomplete or difficult to read.
In-use model adaptation
Design methods that allow models to adjust themselves to new formats, different fonts, or variations in scan quality during use. The goal is to make the models more robust when dealing with variation across documents being processed and changes in these documents over time.
Effective adaptation of vision-language models
Take large AI models that combine image and text analysis and adapt them to specialized document-related tasks, with few annotated examples or from unannotated data. This work aims to reduce manual data preparation and computational resource needs.
Explainable artificial intelligence for secure and transparent document processing
Develop methods that highlight elements that have influenced a result, such as inconsistency in a date, signature, or layout. These methods strive to make AI results more understandable to support human verification and to foster trust in these tools.
Research impact
Ultimately, the methods developed could be integrated into IMDS solutions used in the financial, government, and insurance sectors. They would help strengthen the reliability of digital services and confidence in the documents on which many decisions are based.
Less manual verification
Reducing the time spent processing and manually verifying documents.
Sharper fraud detection
Detecting forgeries and tampering more effectively, especially when facilitated by generative AI.
The next generation of applied AI specialists
Giving students and postdoctoral fellows research experience closely linked to industry needs.
Our industrial partner
IMDS is a Montreal-based company specializing in image processing, intelligent document analysis, and digital identity solutions. Specifically, it develops character recognition, data extraction, document verification, and facial recognition technologies. Its clients include major government bodies, banks, and insurance companies in Canada, Europe, and Africa.
As part of the Chair, IMDS will contribute expertise and tools, as well as data representative of actual situations. The company will participate in validating the methods developed, integrating these methods into its solutions, and training teams.
Our Team
The Chair brings members of ÉTS’ Imaging, Vision and Artificial Intelligence Laboratory (LIVIA) together with IMDS representatives.
At ÉTS
At IMDS
- Patrick Gautschi
Chairman of the Board of Directors - Frédéric Cornu
President - Nicolas Loiseau
Vice President, Research and Development
Contact the Chairholder
Eric Granger
Professor – Systems Engineering Department
1130, rue William
Local F-5088
Montreal, QC H3C 1K3