Singapore's Public Databases Struggle With Duplicate Image Crisis

Singapore's push to digitise everything — government services, property listings, medical records, heritage archives — has produced an unintended side effect: hundreds of thousands of duplicate and misplaced images sitting inside databases that agencies and private platforms depend on daily. The problem is not new, but the scale of it has grown quietly for over a decade, and the systems now being built to fix it reflect just how far the rot had spread before anyone moved decisively.

The issue matters now because Singapore's Smart Nation ambitions rest on data integrity. The Infocomm Media Development Authority has spent several years pushing agencies toward integrated data pipelines. When images attached to the wrong record — say, a Clementi HDB block photograph filed under a Woodlands address, or a patient photograph duplicated across multiple clinic entries in the National Electronic Health Record system — those errors propagate downstream into dashboards, AI training sets and public-facing portals. A duplicated image is not merely an aesthetic inconvenience. It is a data provenance failure.

How the Backlog Built Up

The story begins roughly around 2010, when multiple Singapore government agencies accelerated their moves away from paper filing. The then-National Archives of Singapore, now operating under the National Library Board on Victoria Street, began ingesting digitised photographs at scale. At the same time, the Housing and Development Board was uploading unit photographs for its resale flat portal, and restructured hospitals under the Ministry of Health were linking patient images to electronic records. Each system had its own metadata standard. None of them talked to the others.

Contractors handling digitisation projects were typically paid per record processed, not per record verified. That payment structure created an incentive to upload fast and flag errors later. In practice, flagging rarely happened. When agencies merged databases or migrated to new platforms — as happened repeatedly between 2015 and 2020 as government systems consolidated under the Government Technology Agency, or GovTech — duplicate images from legacy systems were carried across. Error rates compounded rather than cleared.

PropertyGuru and 99.co, the two dominant private property listing platforms in Singapore, encountered a version of the same problem as they aggregated agent-uploaded content. Agents photographing units at Toa Payoh or Bishan would upload images from previous listings by mistake, or a system sync error would attach an old photograph to a freshly listed unit. By 2022, both platforms had introduced automated duplicate-detection tools using perceptual hashing — a technique that compares image fingerprints rather than pixel-by-pixel content. The technology existed. The will to mandate it across public-sector databases took longer to develop.

The Regulatory and Technical Response

GovTech formally acknowledged the duplicate image problem in its 2023 Digital Government Blueprint progress update, identifying image deduplication as one of several data hygiene priorities for the following two financial years. The agency has since piloted a centralised asset registry for government-held digital images, with the National Library Board and the Housing and Development Board among the early participants. Under the pilot, each image is assigned a unique hash on ingest, and any subsequent upload that matches an existing hash triggers a human review workflow rather than automatic acceptance.

The cost of remediation is not trivial. Government technology projects in Singapore have historically run at between S$2 million and S$20 million for mid-scale platform overhauls, depending on scope. Deduplication at the scale of a national image repository — the National Archives holds upward of two million digitised items — requires both computational infrastructure and curatorial labour. The National Library Board has been recruiting digital preservation specialists, with positions advertised on Careers@Gov as recently as March 2026.

For residents and businesses dealing with the practical fallout — wrong flat photographs on HDB's resale portal, misfiled clinic images, heritage records with duplicated metadata — the immediate advice from GovTech's helpdesk is to report discrepancies through the Singpass Feedback portal, which routes submissions directly to the owning agency. GovTech has said the centralised registry pilot is expected to extend to five additional agencies by the end of the 2026 financial year, meaning the full fix is still some distance away. The architecture for preventing future accumulation is being built. Clearing what already exists is the longer, harder task.