How Singapore's Image Duplication Problem Became a Policy Headache: The Road to Duplicate Image Replacement

Singapore's digital housekeeping problem has a name now. Across government portals, public housing databases and municipal websites, duplicate images — the same photograph filed under multiple records, sometimes hundreds of times — have been quietly degrading data quality for the better part of a decade. The Housing and Development Board, the Urban Redevelopment Authority and several statutory boards are now in the process of rolling out duplicate image replacement protocols, a systematic effort to audit and clean visual records that have grown bloated and inconsistent since the mass digitisation push of the mid-2010s.

The timing is not accidental. Singapore's ambitions as a regional artificial intelligence and data hub — anchored in part by the National AI Strategy 2.0 framework released in 2023 — depend on clean, reliable datasets. Duplicated or mislabelled imagery undermines the machine-learning pipelines that government agencies are increasingly using for everything from flat condition assessments in Queenstown and Toa Payoh to infrastructure monitoring along the Pan Island Expressway corridor. Garbage in, garbage out is not a metaphor the Smart Nation and Digital Government Office takes lightly.

How the Duplication Built Up

The roots of the problem go back to around 2015 and 2016, when multiple agencies independently accelerated their shift from paper records to digital asset management systems. The process was fast and, in many cases, siloed. A photograph of a Bishan estate block, for instance, might have been uploaded separately by a town council, by HDB's own estate management division and by a third-party contractor conducting a structural survey — each filing it under a different record identifier, none aware of the others. Over years, with resale flat listings on the HDB Resale Portal and property entries on OneMap accumulating fresh uploads without systematic deduplication, the redundancy compounded.

By the early 2020s, internal audits — details of which have not been made public — reportedly flagged that a meaningful share of image assets held across several agencies were near-identical duplicates. The problem was not merely aesthetic. Duplicate records inflate storage costs, slow search indexing and, critically, introduce inconsistencies when agencies share data. An image tagged with one set of metadata in one system and different metadata in another creates downstream errors when those records are merged or cross-referenced.

The Government Technology Agency, known as GovTech, has been the coordinating body for the remediation effort. Its Data Quality framework, which GovTech has described publicly as a pillar of Singapore's data governance posture, provides the policy scaffolding under which duplicate image replacement sits. The work is technically unglamorous — running perceptual hashing algorithms against image libraries, flagging near-matches for human review, then replacing or retiring redundant files — but the scale is significant given how many public-facing platforms are involved.

What the Clean-Up Looks Like on the Ground

Citizens interacting with the HDB Resale Portal or checking property listings on OneMap may notice, over the coming months, that some estate photographs are refreshed or that older, lower-resolution images give way to current ones. In areas like Tampines and Buona Vista, where large-scale Build-To-Order projects have changed the streetscape substantially since the mid-2010s images were first filed, the visual update alone will be a meaningful accuracy improvement.

For businesses, the practical upshot is more direct. Property technology firms that licence data feeds from government sources have long complained informally that image quality and consistency across records is uneven. A cleaner image database makes automated valuation models and listing aggregators more reliable.

The effort also connects to broader cost pressures. Cloud storage is not free, and with Singapore's public sector cloud expenditure rising steadily under various digital government initiatives, eliminating redundant files carries a real, if modest, financial logic alongside the data integrity argument.

Agencies have not published a completion timeline for the full deduplication exercise, but GovTech's broader data quality initiatives are understood to be ongoing through at least the end of 2026. For residents and businesses that rely on government digital services, the practical advice is straightforward: if an image or record on a public portal looks outdated or inconsistent with what you know to be current, use the official feedback channels — HDB's e-feedback portal and OneMap's report function both accept image correction submissions — rather than assuming the record is authoritative.