Paris Takes Aim at Duplicate Images Flooding Its Public Record — But Is It Moving Fast Enough?

Paris has a redundancy problem buried inside its own digital infrastructure. Thousands of duplicate photographs — many of them official images of public works, heritage sites, and city planning projects — are cluttering the databases managed by the Bibliothèque nationale de France and the city's Direction de l'Urbanisme, making records harder to search, verify, and archive. Librarians and urban planners quietly flagged the issue earlier this year, and the Mairie de Paris confirmed in a June 2026 internal review that the problem had grown substantially since the post-Olympics documentation surge that followed the Paris 2024 Games.

The timing matters because France is mid-cycle on several enormous public investment projects. The Grand Paris Express, the largest urban transport project in Europe by track length, has generated hundreds of thousands of site photographs since groundbreaking. Société du Grand Paris, the public body overseeing the metro expansion, relies on that visual record for procurement audits, contractor accountability, and public transparency reporting. When the same image appears under multiple file identifiers, it creates audit gaps that cost money and time to untangle. The Paris 2024 legacy activation programmes — particularly the Seine riverside regeneration between the Pont de Bercy and the Pont d'Iéna — have compounded the problem by adding a second wave of overlapping documentation to already strained city servers.

What Paris Is Doing — And Where It Lags

The city is not starting from scratch. The BnF has operated its Gallica digitisation platform since 1997, and the archive contains more than eight million documents. In 2025, the institution launched a machine-learning deduplication pilot covering roughly 400,000 images in the contemporary urban photography collections. The pilot, developed in partnership with the Institut national de recherche en sciences et technologies du numérique — better known as Inria — uses perceptual hashing to flag near-identical images even when file names, metadata, or compression rates differ. Early results, presented at a February 2026 conference in the 13th arrondissement, suggested the tool correctly identified duplicates at a rate above 90 percent in test batches.

That is a creditable start. But Amsterdam's Stadsarchief completed a comparable project for its entire municipal photography collection in 2023, clearing roughly 1.2 million images and reducing server storage costs by around 18 percent, according to figures the archive published at the time. Tokyo's National Diet Library finished a phased deduplication programme across its digital map and aerial photography holdings in late 2024, integrating the cleaned records directly into its public-facing NDL Digital Collections portal. Paris, by contrast, has not yet committed to a full rollout date for the BnF pilot beyond the test phase, and the Direction de l'Urbanisme is running a separate, incompatible system with no announced plans to merge the two efforts.

That fragmentation is the core issue. The Hôtel de Ville has three distinct departments — culture, urban planning, and transport coordination — all maintaining image databases with different metadata standards. A photograph taken at the Gare du Nord renovation site in March 2025 could exist in all three systems under different file identifiers, with no automated flag linking them. This is not a theoretical problem: project managers working on the Boulevard Périphérique transformation plan reported earlier this year that cross-referencing visual documentation between departments added weeks to their compliance reporting cycles.

What Happens Next

A working group convened by the Préfecture de Paris is expected to publish recommendations before the end of the third quarter of 2026. The group reportedly includes representatives from Inria, the BnF, and Société du Grand Paris, though no official membership list has been released. If the recommendations lead to a unified deduplication standard — something Amsterdam achieved through its Digitaal Erfgoed programme and Tokyo through a national library mandate — Paris would likely need 18 to 24 months for full implementation across all city-managed databases.

For residents and researchers trying to navigate public planning records right now, the BnF's Gallica portal remains the most functional starting point. The Direction de l'Urbanisme's Atlas de Paris platform, accessible via the Mairie de Paris website, also allows image searches by arrondissement. The practical advice from archivists familiar with both systems: always cross-reference any image found in one database against the other, and treat file creation dates with particular scepticism — metadata inconsistencies are common, and the deduplication work is far from finished.