How Singapore's Digital Archives Ended Up Full of the Same Image Twice — and What's Being Done About It

Singapore's public sector agencies and cultural institutions are sitting on a quiet but costly problem: thousands of duplicate digital images, stored redundantly across servers maintained by bodies ranging from the National Library Board to the Urban Redevelopment Authority, inflating storage costs and complicating public access to records that Singaporeans increasingly rely on for everything from heritage research to property transactions.

The issue matters right now because the government's push to consolidate digital infrastructure — accelerated under the Smart Nation and Digital Government Office's ongoing GovTech-led programmes — has brought auditors face to face with legacy repositories that were built in silos. Each agency digitised its own photo archive independently, often during the early 2000s, and many did so again when higher-resolution scanning became affordable in the 2010s. Nobody was keeping a centralised ledger of what already existed.

How the Duplication Built Up

The roots of the problem trace back to at least 2001, when the National Archives of Singapore launched its first large-scale digitisation push at its repository along Canning Rise. The project was a genuine achievement — hundreds of thousands of historical photographs, maps and documents were scanned and indexed. But other agencies were running parallel efforts. The National Heritage Board digitised collections held at the Asian Civilisations Museum on Empress Place. The Housing and Development Board scanned its own photographic records of public housing construction going back to the 1960s. The Sentosa Development Corporation archived island-development imagery separately.

When any two agencies happened to hold physical copies of the same photograph — which was common for images of national events, state visits and major infrastructure openings — both agencies digitised their copy independently. The result was functionally identical image files living on different servers, tagged with different metadata, sometimes in incompatible formats. Researchers using the National Library Board's digital portal, NL Online, and the National Archives' digital platform, Archives Online, would occasionally encounter the same photograph twice without realising it came from duplicated holdings.

The problem compounded through social media-era re-uploads. Public agencies routinely shared images on government websites, the Singapore Tourism Board's promotional channels and official Facebook pages, after which the images were sometimes re-ingested into archival systems during subsequent digital preservation sweeps — creating third and fourth copies of a single file.

The Cost and the Fix

Cloud storage is not free. GovTech's Government Commercial Cloud migration, which shifted agency workloads to Amazon Web Services and Microsoft Azure beginning around 2018, exposed the true scale of redundant image data because cloud billing is volume-sensitive in a way that on-premises servers were not. Agencies found themselves paying for storage of files that contributed no informational value beyond what a single copy would provide.

The practical response has been the adoption of perceptual hashing — a technique that generates a short fingerprint for each image based on its visual content rather than its file name or metadata. Two photographs of the Merlion at Fullerton Waterfront taken on the same day, scanned at different resolutions and labelled differently, will still produce near-identical perceptual hashes and can be flagged as probable duplicates for human review. GovTech began piloting this approach within selected agency repositories in 2023.

The challenge is governance, not technology. Deciding which copy of a duplicate image is the canonical version — and therefore which metadata record survives — requires inter-agency coordination that has historically been slow. The Infocomm Media Development Authority has been working on data interoperability standards that could give agencies a common framework for these decisions, though implementation timelines remain tied to broader whole-of-government data architecture reviews.

For members of the public who use archives regularly — genealogists, students at Nanyang Technological University, journalists, property researchers pulling historical images of Toa Payoh or Queenstown developments — the practical takeaway is straightforward. Duplicate-replacement projects will temporarily take certain image collections offline for reindexing. The National Archives has advised users to download materials they need for active projects before any scheduled maintenance windows, details of which are published on the Archives Online portal at nas.gov.sg. The inconvenience is real but finite. The alternative — leaving duplicate bloat to grow unchecked — costs more in every sense.