The document analysis and recognition domain is a foundational and historical area of expertise of the Image Processing and Pattern Recognition group. Historically, this domain provided real-world challenges and served as a way to validate the mathematical morphology methods developed by the team. Over the past few years, the team’s work has increasingly integrated natural language processing components into traditional computer vision tools. We also engage with traditionally separate communities such as Digital Humanities as well as Human and Social Sciences and maintain collaborative relationships with researchers in these fields. These partnerships provide us concrete use cases and development needs.
Our contributions in document analysis and recognition
Rooted in our robust expertise in efficient image processing, we contribute to a wide range of challenges in document analysis and recognition, including:
- Low-level shape recognition [2][3]
- Image classification and recognition
- Layout analysis
- Image-to-text and image-to-structure systems
- Extraction of structured information from textual content [4]
- Data linking and anomaly detection
- Visualization and collaborative annotation
Over the year, the team has contributed to numerous industrial and academic collaborative projects on a broad range of document modalities ranging from administrative and identity documents (MOBIDEM project) to historical document collections (ANR SoDUCo, MEZANNO project).
The team has also been involved with the organization of the ICDAR 2021 MapSeg Challenge [1] and ICDAR 2024 MapText Challenge [5]
Related Projects
AGODA
AGODA is a digital humanities project funded by the DataLab of the Bibliothèque nationale de France. It aims to make the parliamentary debates of the French Third Republic (1881–1940) more accessible and usable by creating a structured, semantically enriched XML-TEI corpus.
MEZANNO
The MEZANNO project develops open, AI-assisted tools to support the construction, extraction, and semantic structuring of customized textual corpora from digitized archival documents using the IIIF standard.
MOBIDEM
The MOBIDEM project aimed to make secure and user-friendly mobile electronic signatures broadly accessible by developing a platform for identity verification and signature issuance without in-person appointments.
SoDUCo
The ANR SoDUCo project (2019–2023) aimed to build an open historical database of Paris by extracting and analysing urban and social data from maps and trade directories spanning 1789 to 1950.
Related Publications
[1]
Joseph Chazalon • Edwin Carlinet • Yizi Chen • Julien Perret and Bertrand Duménieu • Clément Mallet • Thierry Géraud • Vincent Nguyen • Nam Nguyen • Josef Baloun • Ladislav Lenc • Pavel Král. "ICDAR 2021 Competition on Historical Map Segmentation". Proceedings of the 16th International Conference on Document Analysis and Recognition (ICDAR'21). 2021. https://doi.org/10.1007/978-3-030-86337-1_46.
[2]
Yizi Chen • Edwin Carlinet • Joseph Chazalon • Clément Mallet and Bertrand Duménieu • Julien Perret. "Vectorization of Historical Maps Using Deep Edge Filtering and Closed Shape Extraction". Proceedings of the 16th International Conference on Document Analysis and Recognition (ICDAR'21). 2021. https://doi.org/10.1007/978-3-030-86337-1_34.
[3]
Philippe Bernet • Joseph Chazalon • Edwin Carlinet • Alexandre Bourquelot • Élodie Puybareau. "Linear Object Detection in Document Images Using Multiple Object Tracking". Proceedings of the International Conference on Document Analysis and Recognition (ICDAR 2023). 2023. https://doi.org/10.1007/978-3-031-41734-4_28.
[4]
Solenn Tual • Nathalie Abadie • Joseph Chazalon • Bertrand Duménieu • Edwin Carlinet. "A Benchmark of Nested Named Entity Recognition Approaches in Historical Structured Documents". Proceedings of the International Conference on Document Analysis and Recognition (ICDAR 2023). 2023. https://doi.org/10.1007/978-3-031-41682-8_8.
[5]
Li, Zekun • Lin, Yijun • Chiang, Yao-Yi • Weinman, Jerod • Tual, Solenn • Chazalon, Joseph • Perret, Julien • Duménieu, Bertrand and Abadie, Nathalie. "ICDAR 2024 Competition on Historical Map Text Detection, Recognition, and Linking". Document Analysis and Recognition - ICDAR 2024. 2024. https://doi.org/10.1007/978-3-031-70552-6_22.