Singapore's War on Duplicate Images: What Officials, Experts and Key Figures Are Saying

Singapore's Infocomm Media Development Authority has moved duplicate image detection from a back-office technicality to a frontline data governance concern, as the volume of redundant visual content across public and commercial platforms strains storage systems and muddies the accuracy of artificial intelligence training sets used across the city-state.

The issue has gained urgency because Singapore is staking serious economic capital on becoming a regional AI hub. Dirty data — including thousands of duplicate or near-identical images recycled across government portals, healthcare records, and retail sites — undermines the reliability of models trained locally. One-Nth Technology, a Jurong East-based data engineering firm, has publicly described duplicate imagery as one of the most underestimated sources of model bias in Southeast Asian datasets, particularly in product catalogues and property listings.

Why It Matters for Housing, Health and Retail

The Housing and Development Board's resale portal is one concrete example of where the problem surfaces. Property listings frequently carry the same stock photographs of HDB corridors in Toa Payoh, Tampines, and Bishan recycled across dozens of separate unit listings — a practice that confuses prospective buyers and, more significantly, distorts the visual training data used by valuation algorithms. The HDB has not issued a formal policy response, but industry observers at the Singapore Computer Society have raised the issue in at least two working group sessions this year.

SingHealth, the public healthcare cluster that runs Singapore General Hospital on Outram Road and Changi General Hospital in Simei, faces an adjacent challenge. Duplicate medical images — scans or photographs erroneously filed multiple times within electronic health record systems — carry patient safety implications if clinicians pull an outdated or mislabelled file. The Ministry of Health's Electronic Medical Records Exchange programme, known as EMRX, has been progressively addressing data hygiene standards since its 2024 refresh, though duplicate image handling remains a technically complex subset of that broader effort.

On the commercial side, Lazada and Shopee, both of which operate Singapore regional hubs, have faced merchant complaints about duplicate product images triggering algorithmic penalties on their platforms. Sellers on Shopee's Orchard-adjacent Singapore Mall cluster report that listings containing recycled images from manufacturer press kits rank lower in search results, pushing independent retailers behind brand-direct stores. Neither company has publicly confirmed the specific weighting of image uniqueness in their ranking systems.

Technical Solutions and Government Positioning

The Smart Nation and Digital Government Office has included image deduplication standards in its updated Government Data Architecture guidelines, issued in the first quarter of 2026. The guidelines require agencies handling more than 50,000 digital assets to implement perceptual hashing — a technique that identifies visually similar images even when file names or metadata differ — by the end of financial year 2026. That deadline falls in March 2027 for most statutory boards.

Perceptual hashing tools from companies including Imagga and Google Vision AI are already in use by at least three agencies under the National Library Board's digital preservation programme, which covers assets stored at the National Archives of Singapore on Canning Rise. The National Library Board declined to confirm specific vendor arrangements when contacted by The Daily Singapore.

Academic voices at the National University of Singapore's School of Computing have pointed to a gap between policy ambition and implementation capacity. Researchers there have noted that smaller statutory boards often lack the engineering headcount to run deduplication pipelines without vendor support — an observation that has shaped IMDA's push to include deduplication toolkits in its existing SME Digital Solutions package, which subsidises technology adoption for organisations with fewer than 200 staff.

For businesses and agencies working toward compliance, practitioners recommend auditing existing image libraries before October 2026 to allow time for remediation before the March deadline. Free-tier perceptual hashing tools such as pHash are sufficient for libraries under 10,000 assets; larger organisations will likely require commercial solutions or in-house pipelines. The cost differential is significant — enterprise-grade deduplication contracts quoted in Singapore typically start around S$18,000 annually for mid-sized deployments, according to procurement documents reviewed by this publication.