How Paris's Public Image Archive Ended Up Drowning in Duplicates — and What the City Is Doing About It

Paris's municipal image library holds more than 2.4 million files. A significant portion of them — city archivists have been working since late 2025 to quantify exactly how many — are duplicate or near-duplicate photographs, the accumulated residue of decades of poorly coordinated acquisition across a dozen separate city directorates. The problem is not new, but the cost of ignoring it finally is.

The reckoning arrived because of scale. The Paris 2024 Olympic and Paralympic Games generated an extraordinary volume of official documentation photography — tens of thousands of images commissioned by the Ville de Paris, the Préfecture de Police, the Comité d'Organisation des Jeux Olympiques, and various legacy bodies, often covering the same venues on the same days. When those files were folded into the central repository managed by the Direction des Affaires Culturelles on the Rue des Francs-Bourgeois, staff discovered that deduplication protocols that had been adequate for routine municipal photography simply could not handle the volume. Server costs at the city's primary data centre climbed. Retrieval times slowed. And journalists and urban planners trying to access images of, say, the Seine-Saint-Denis competition venues found themselves wading through hundreds of near-identical shots before locating usable material.

A Problem Built Over Decades

The roots of the duplication crisis run back further than the Olympics. From the early 2000s onward, Paris's administrative culture encouraged individual departments — urban planning, tourism, heritage, social affairs — to commission their own photography rather than draw from a shared pool. The logic at the time was operational: a department needed images on its own timeline and did not want to wait for central procurement. The Apur urban planning agency, the Office du Tourisme et des Congrès de Paris, and the Direction de l'Urbanisme each built their own partial archives. Cross-referencing between them was minimal.

Grand Paris Express construction documentation alone, running since the first ground-breaking works in 2016, has produced imagery from at least six separate contracted photographers covering overlapping stretches of the same worksites — stations at Saint-Denis Pleyel, Bagneux, and Villejuif among them. Sources familiar with the archive structure, speaking in general terms about the institutional problem rather than specific contracts, describe a situation where the same infrastructure trench might appear under three different file taxonomies in three different departmental folders, none flagged as related to the others.

Cloud storage is not free. The Ville de Paris signed a multi-year infrastructure contract in 2022 valued at several tens of millions of euros covering data hosting across municipal services. Imagery represents a disproportionate share of that storage burden because photographic files are large. Reducing the duplicate load by even 20 percent would carry measurable savings — and more importantly, would make the archive genuinely usable rather than nominally complete.

What the Deduplication Programme Looks Like in Practice

Since October 2025, a working group convened under the Direction des Affaires Culturelles has been piloting automated hash-matching software on a subset of the collection — starting with images tagged to arrondissements 18, 19, and 20, areas heavily documented during both the Olympic torch relay and ongoing Seine-Saint-Denis regeneration programmes. The pilot covers roughly 180,000 files. Human reviewers then assess flagged near-duplicates, because automated tools cannot reliably distinguish between two photographs of the Sacré-Cœur taken seconds apart and two photographs that happen to be compositionally similar but document different events.

The broader rollout is planned in phases through 2027, with the complete archive scheduled for review before the Ville de Paris's next municipal digital strategy cycle begins. Departments are being asked — some for the first time — to agree on shared metadata standards, so that a photograph of the Canal de l'Ourcq taken for a tourism brief and one taken for a flood-risk planning document can be cross-referenced rather than siloed.

For Parisians and researchers who rely on the public-facing Bibliothèque historique de la Ville de Paris on the Rue de Rivoli, the practical upshot should eventually be a cleaner, faster image search. That outcome is still some way off. For now, the work is unglamorous: comparing checksums, reconciling taxonomies, and deciding, file by file, what the city actually needs to keep.