Project duration: 2018 - 2021 Date coverage: 1920 - 2021 Website Contact

Photos Platform for data aggregation or retrieval Machine Learning Tools for data mining/extraction Linked Data 3D / 4D modelling and reconstruction Archive Geohistory Digital Humanities Digital Cultural Heritage Digital Tourism

ALEGORIA: Advanced Linking and Exploitation of diGitized geOgRaphic Iconographic heritAge

ALEGORIA aims at facilitating the promotion of iconographic institutional collections describing the French territory in various periods going from the interwar period to our days. The collections in sight are of variable sizes, between thousands and hundred thousand elements, and are constituted by iconographic representations, which are most of the time oblique, vertical aerial or terrestrial photographs of the environment. Contrary to the well anchored exploitation of the satellite imagery where the professional practices are many (scientists, civilians and militaries) and the data well identified and structured, the promotion of those collections remains confidential and scattered. They are spread “in silos” within various institutions, digitized partially, generally not or little documented and weakly georeferenced. Nevertheless they represent a rich heritage, under-exploited by the general public as well as by professionals. The development of such a heritage would benefit from tools allowing to automate their collection, their processing and their indexing, inside the collection itself but also between collections, to be capable of crossing them and then of studying them better. Their implementation within applications implies to facilitate their apprehension by non-specialist users, it supposes to supply friendly tools for the visual exploration of these collections. Within this context, ALEGORIA has for objectives the proposal of two innovative platforms of demonstration dedicated to the promotion of the dematerialized and digital-born iconographic geographical heritage:

  • A multimodal and large-scale engine for retrieval, coupling search by content and by metadata in intra-domain collections as well as cross-domain collections;
  • A web-based engine, based on the 3D French Geoportal iTowns, for the immersive display of these contents, allowing spatiotemporal navigation and interaction in the 3D environment enriched with ancient photographic collections.

Datasets. One ambition of ALEGORIA is also to make datasets available to the communities for research purposes, in particular:

  • Standardized metadata describing the image collections of the project (RDF / RiCO) and published as linked open data;
  • Large-scale and annotated image datasets dedicated to the training of image descriptors for researchers: SF300 dataset and ALEGORIA benchmark;
  • A multi-date and multimodal dataset dedicated to the description of geographical entities with neighborhood relationships: GENR.

Use cases. To evaluate the tools proposed, several use cases are studied:

  • Evaluation as a tool for the structuring of iconographic heritage collections to better understand the metamorphosis of post-war territories (Archives Nationales – French territory);
  • Valorization of iconographic collections outside the walls for cultural tourism (Nicéphore Niépce Museum – Chalon-sur-Saône);
  • Evaluation as a research and communication tool in urban visual sociology (LAVUE – Greater Paris, Nanterre).

Project partners

Associated partners

Other partners

  • LIRIS - Ecole Centrale Lyon
  • LAVUE - Université Paris Nanterre
  • Musée Nicéphore Niépce

Public datasets


The metadata of the photographs dealt with by the ALEGORIA project are published as Linked Open Data. They describe the following photographic fonds: - Obliques aerial photographs (IGN - Photothèque nationale) - Lapie (Archives nationales) - Ministère de la Reconstruction et de l'Urbanisme (Archives nationales) - Combier (Musée Nicéphore Niépce) A short presentation of these collections can be found here:

License: Custom license, please specify

ALEGORIA-DB image benchmark

The ALEGORIA-DB benchmark is an image dataset involving heterogeneous geographical heritage images of various objects of interest in French urban and natural scenes, through a time period ranging from the 1920s to nowadays. The content is highly characterized by multi-date, multi-source and multi-view images. ALEGORIA-DB consists of a total of 13175 images of high resolution (800px*variable), decomposed as follows: 1859 query images dispatched in 58 classes, where each class is defined around an object or a location in urban or natural scenery ; 11316 relevant image distractors. One originality of the dataset is that each query image was manually annotated with several quantized attributes associated with image variations, making it relevant for the evaluation of approaches facing these variations.

License: Custom license, please specify

GENR: Geographical Entities with Neighborhood Relationships

Geographical entities with neighborhood relationships : these datasets are excerpts of the reference vector database produced by the French Mapping Agency (IGN), namely the BDTOPO®. Associated with them are graphs representing the geographical entities present on a 200 m x 200 m area, defined around a given point of interest, as well as their neighborhood relationships. Each dataset represents a French administrative unit (Moselle, Bas-Rhin, Meurthe-et-Moselle, Côtes-d'Armor), at a given period (2004, 2010, 2014, 2019).

License: Custom license, please specify

SF300 large-scale image dataset

The SF300 dataset is a large-scale dataset dedicated to the training of deep descriptors for remote sensing tasks. With orientation variations and a class-definition based on geolocation rather than task-specific annotations, it is particularly suited for fine-tuning CNN feature extractors for remote sensing classification or retrieval. The dataset was made using images taken from a plane and covers a wide range of landscapes on the Danish territory. It consists of 308k images in 27.5k unique geolocations (classes) for the train set, and 21.8k images in 2.4k classes for the test set that can be used as the val split when fine-tuning. Images are all 512x512.

License: Custom license, please specify

With financial support by