Istanbul's Digital Archive Push Stumbles Over Duplicate Image Crisis — Here's What Happened This Week

Thousands of duplicate photographs — some estimates put the figure at roughly 40,000 redundant image files — have accumulated across Istanbul's two main heritage digitisation platforms since January, administrators working on the project confirmed this week. The problem surfaced publicly on Tuesday when the Istanbul Metropolitan Municipality's Urban Transformation Directorate flagged the backlog in an internal technical report circulated to partner institutions, according to documents reviewed by The Daily Istanbul.

The timing is awkward. The municipality is midway through a five-year, 2.3 billion lira digital inventory programme launched in 2024, designed to catalogue every structurally at-risk historic building in districts like Fatih, Balat and Ayvansaray before updated seismic codes come into force at the end of 2026. Earthquake risk has sharpened since the February 2023 Kahramanmaraş disaster. Getting the photographic record right is not an administrative footnote — engineers use it to compare before-and-after structural states when fast-tracking demolition or reinforcement orders.

Where the Duplication Happened

Two systems are at the centre of the problem. The first is the municipality's own Istanbul Kent Rehberi platform, which hosts geo-tagged building photographs contributed by field surveyors across 39 districts. The second is the digital repository maintained by Istanbul Technical University's Faculty of Architecture, located on the Taşkışla campus in Maçka, which has a data-sharing agreement with the municipality signed in March 2025. Both systems ingested the same batch of survey images shot during a joint fieldwork sprint in Süleymaniye and Zeyrek in late April, and neither platform's automatic deduplication filter flagged the overlap at the time.

ITU's repository alone grew by roughly 18,000 files during that April sprint. Officials now say somewhere between 30 and 45 percent of those uploads are exact or near-exact duplicates, depending on which similarity threshold the filter applies. Storage is not the primary concern — cloud capacity costs the project approximately 4,200 lira per terabyte per month under its current contract. The real cost is human: trained cataloguers at the Fatih district office on Macar Kardeşler Caddesi have spent much of the past week manually reviewing flagged images rather than processing new survey zones.

The deduplication failure also touches a broader tension in Istanbul's heritage sector. Preservationists have long argued that Balat and Fener, the adjoining Byzantine-era neighbourhoods along the Golden Horn, receive disproportionate photographic coverage compared to inland areas like Karagümrük or Cerrahpaşa, where structural risk is equally high but tourist footfall is lower. Duplicate images from well-photographed streets in Balat effectively crowd out indexing capacity for less-documented sites.

What Comes Next

The Urban Transformation Directorate has given its technical partners until July 18 to implement a hash-based deduplication protocol that should automatically reject identical file signatures at the point of upload. A secondary perceptual-hash layer — designed to catch near-duplicates shot from slightly different angles — is expected to follow in August, though that timeline depends on procurement approval for the software licence, which carries a quoted cost of around 280,000 lira.

In the interim, cataloguing work on the Zeyrek zone — a UNESCO World Heritage buffer area — has been paused while supervisors clear the backlog. Field survey teams have been redirected this week to Eyüpsultan and Gaziosmanpaşa, two districts north of the Golden Horn that have their own seismic vulnerability assessments pending before October.

For residents and property owners in affected neighbourhoods, the practical advice from the directorate is straightforward: structural assessment appointments already scheduled through the municipality's ALO 153 line remain unaffected. The image-database problem sits upstream of those assessments and does not delay individual building reports already in progress. Anyone who has not yet registered a structurally at-risk property can still do so through the e-Devlet portal or in person at district municipality offices — the July 18 technical deadline does not affect that process.

The broader programme remains on schedule, for now. But with five months left before the new seismic code deadline, the margin for further technical disruption is thin.