Artificial intelligence for the stained glass windows of Notre Dame Cathedral in Chartres


The artificial intelligence application will be designed for iOS and Android smartphones and tablets. It will enable visitors to discover Chartres Cathedral and its finest stained glass windows in high definition.
Completely free of charge, it will be available in three languages: French, English and German.

The application will meet a dual challenge:

  • refocusing the discovery of the cathedral on its most beautiful stained-glass windows, as well as lesser-known ones
  • promote discovery of the associated museum.

Two concrete examples:


It’s not always easy for children to discover the cathedral. Stained-glass windows remain abstract, as the details are not very visible.

In 2020, 92% of children own a cell phone, and they spend most of their time playing games or chatting on social networks.

By having children download the application, teachers and parents will have a fun tool to help them discover the stained-glass windows. They can even ask them to search for details in the form of a game.

Our visitors

A guided tour is not always possible for many visitors. In fact, with the app, they can organize their own itineraries and understand the history of the stained glass windows that make the biggest impression on them.

AI functional overview

Step by step, the user will be able to perform the following operations thanks to AI:

  1. Take a photo of the stained-glass window of interest using the smartphone application;
  2. The photo is recognized by the AI, which processes the information with the media center.
  3. The photo of the stained-glass window is displayed with a text underneath explaining the stained-glass window.
  4. Additional, longer text can be added by clicking on a “Learn more” button.
  5. On the photo of the stained-glass window, points of interest (POIs) are indicated for the user to click on.
  6. The POI may correspond, for example, to a medallion, a decorative element, a character…
  7. In this case, a page opens and the user can read the explanation corresponding to the element.
  8. The photographed stained-glass window is saved on the user’s phone for later viewing.

Its strengths:

A unique solution for discovering the history of what you see;

An exclusive solution for real-time recognition of stained glass windows, based on artificial intelligence (AI), to enable visitors to match their photograph of any stained glass window to a database, specially created to retrieve all the information relating to each stained glass window in the cathedral;

A high-performance AI that will continually evolve: the result of R&D and the latest research in Deep Learning, the solution is specially trained to recognize the stained-glass windows photographed, regardless of meteorological constraints (cathedral bathed in light or darkened by a storm), shooting conditions (shot from an angle, from below, from a greater or lesser distance, etc.) and the quality of the lenses and zooms used. The software will thus be continually enriched, to integrate the ever-increasing diversity of smartphones and tablets on the market;

A stained-glass database that will gradually be expanded over the months, until it includes all the cathedral’s stained-glass windows: the application is based on a database initially comprising a selection of some twenty remarkable stained-glass windows, in high definition, and which will gradually be enriched by the extraordinary collection of 350 stained-glass windows on the site, revealing the full mastery of the craftsmen and providing access to additional information, edited by a team of art history specialists.

A service platform reserved for historians and the museum to progressively feed the database and open it up to the public (see “content” paragraph): the platform will make it easy to access, edit and save the data and works collected, and to select those that will be made accessible to cathedral visitors via the application. In this way, the platform will be able to index the stained-glass heritage and eventually serve the entire digital system, from the cathedral application to the Chartres museum(s).

Présentation de l'application d'intelligence artificielle sur les vitraux

In its first version, the application features a menu consisting mainly of the following items:

Practical information for visiting the cathedral: presentation, address, opening times, availability of guided tours

Practical information for visiting the Musée du Vitrail: presentation, address, opening hours, access map, link to the museum website.

Discover stained glass with automatic window recognition
Visit the cathedral with the help of an interactive map, to navigate from one point of interest to another and find your bearings in space.

A technical approach based on the latest technologies, guaranteeing reliable operation. The application will be developed using Ionic technology, a multi-platform “hybrid” solution enabling applications to be deployed on both iOS and Android, from a single code base.

The core of the application will be developed in React and Typescript, the two most widespread hybrid development technologies.

Particular attention will be paid to guaranteeing the high quality of the stained glass images, while limiting their weight in the application, so as to keep the application’s memory size within the norm for the mobile application market, and not to hinder download by visitors (who might consider it too heavy, or be blocked when downloading for lack of space).

Focus on Deep Learning

Deep Learning (or machine learning) is part of the large family of technologies linked to artificial intelligence. It is a branch of algorithmics in which some of the operations listed are not written by the programmer, but result from a phase of automated statistical observation of a data set: learning (or training).

Stained-glass window recognition is based on a Deep Learning solution dedicated to images, designed to train a calculation function between points taken in photographs and reference points.

This time-consuming function will be trained prior to use by visitors to the cathedral, with examples of shots taken from a representative sample of photographs.

The Deep Learning function will then be executed directly in the in situ application to enable recognition of the stained glass. This will be done without the need for an Internet connection to computing servers.