Singapore's Duplicate Image Problem: The Key Decisions That Will Shape What Comes Next

Singapore's push to digitise everything — public housing records, heritage photography, government tender documents, even hawker centre directories — has quietly created a parallel headache: sprawling repositories clogged with duplicate images that inflate storage costs, confuse search results and, in some cases, present conflicting visual records of the same place or project. The question now is not whether to act, but who acts first and how.

The problem has sharpened in 2026 as the Infocomm Media Development Authority accelerates its Smart Nation 2.0 roadmap, pushing agencies toward unified data environments. When multiple government bodies migrate image libraries into consolidated cloud infrastructure, duplicates that once sat silently in separate silos suddenly collide. A photograph of Toa Payoh Hub tagged under three different filenames, or construction-phase images of Tengah's eco-town appearing in duplicate across Urban Redevelopment Authority and HDB project folders, are the mundane but operationally costly results.

Why the Timing Is Urgent

The National Library Board is currently mid-way through a multi-year digitisation drive for the National Archives of Singapore, which holds photographic records dating to the colonial era. Archivists working out of the building on Canning Rise have flagged that automated ingestion tools, while fast, lack reliable deduplication logic sophisticated enough to distinguish a true duplicate from a near-identical image taken seconds apart — a distinction that matters enormously when one version carries handwritten metadata and the other does not.

Meanwhile, GovTech's government commercial cloud migration programme, which has been shifting agency workloads onto AWS and Azure infrastructure since 2023, bills storage by the gigabyte. Duplicated image files are not a trivial line item. Across large portfolios — the Singapore Land Authority alone manages spatial imagery covering the island's roughly 730 square kilometres — redundant files compound into meaningful recurring expenditure that shows up in annual budget submissions to the Ministry of Finance.

The Housing and Development Board's HDB Resale Portal and BTO microsite together carry tens of thousands of unit and precinct photographs. When a block in Bukit Merah or Queenstown is renumbered after a revamp, old images get re-uploaded rather than replaced, and the original files often remain. The result is that prospective buyers searching for a specific block may encounter images from two different renovation cycles without any clear timestamp or label.

The Decisions That Cannot Be Deferred

Three choices are now sitting on agency desks. First, whether to adopt a centralised deduplication layer — essentially a hashing-based filter that flags pixel-identical or near-identical images before ingestion — or leave individual agencies to manage their own libraries. GovTech has the technical capability to build such a layer into the government's data exchange infrastructure; the sticking point is governance, specifically which ministry owns the policy mandate.

Second, replacement protocol. When a duplicate is confirmed, simply deleting one copy is not always appropriate. In archival contexts, the National Heritage Board has noted that two copies of an image may have been annotated differently by different curators, meaning both carry independent informational value. A blanket-delete policy would destroy metadata. A merge-and-retire workflow, where one canonical version absorbs annotation from its duplicate before the secondary file is archived offline, is more defensible but significantly more labour-intensive.

Third, procurement. Agencies that cannot build in-house are looking at commercial image-management platforms. The Government Technology Agency's bulk procurement framework already covers several digital asset management vendors, but contract terms negotiated before the current cloud-migration scale-up may not reflect storage volumes agencies now need to handle.

For residents and businesses that interact with government image systems — contractors uploading project photographs to the Building and Construction Authority's CORENET X portal, for instance — the practical near-term advice is to document every upload with a unique filename convention that includes date, project reference and version number. That one discipline, widely adopted, would blunt the duplication problem at its source. The harder institutional work of retrofitting existing archives falls to the agencies, and the decisions they make in the second half of 2026 will set the operating standard for Singapore's digital government for years to come.