Singapore's Duplicate Image Problem: The Key Decisions Ahead for a City Going Digital

Singapore's public and private digital archives are sitting on a problem that has been quietly compounding for years. Duplicate images — redundant photographs, repeated graphics, and near-identical visual assets stored across multiple servers — are consuming storage capacity, muddying search results, and creating compliance headaches for organisations from the Housing Development Board to Orchard Road retailers managing e-commerce catalogues. The question is no longer whether to fix it. It is who acts first, how, and at what cost.

The urgency has sharpened in 2026. Singapore's push to position itself as a regional AI and data hub — anchored by investments in facilities along Tuas and Jurong Island, as well as the Infocomm Media Development Authority's ongoing Digital Industry Singapore initiative — demands cleaner data infrastructure. Duplicate imagery is not a cosmetic inconvenience. It degrades machine-learning training sets, inflates cloud storage bills, and introduces inconsistencies into automated systems that government agencies increasingly rely on for everything from urban planning to public health monitoring.

Where the Problem Sits and Who Owns It

The duplication crisis spans several distinct domains in Singapore. At the National Library Board's digital preservation unit on Victoria Street, archivists have been grappling with overlapping image records ingested from multiple digitisation drives conducted between 2018 and 2024. The HDB's property portal, which hosts tens of thousands of flat listing photographs submitted by estate agents and owners, has long flagged internal redundancy as a data quality issue — a concern that becomes more pointed as the board accelerates its use of computer vision tools to assess property conditions remotely.

Private sector exposure is significant too. Lazada and Shopee, both operating major logistics and seller-support hubs in the Paya Lebar and Alexandra Road corridors, manage product image libraries running into the hundreds of millions of assets for Southeast Asian markets. Duplicate product photographs across seller accounts inflate catalogue sizes, slow search indexing, and produce inconsistent price-comparison results — a direct hit to consumer trust and platform revenue. Industry estimates from regional technology consultancies have placed the proportion of duplicate or near-duplicate images in large e-commerce catalogues at somewhere between 15 and 30 percent, though precise figures vary widely by platform and product category.

The Decisions That Cannot Wait

Three decisions are now pressing. The first is technical standardisation. Singapore has no single agreed framework for image deduplication across public-sector agencies. The Smart Nation and Digital Government Office has the remit to push cross-agency data standards, but any directive would need buy-in from ministries with their own legacy systems and procurement cycles. A unified hashing and perceptual similarity standard — the technical method by which near-duplicate images are identified — would allow agencies to share deduplication tools rather than each contracting separately.

The second decision concerns data retention law. Singapore's Personal Data Protection Act, last substantially amended in 2020, does not specifically address duplicate visual records. Legal advisers at firms along Raffles Place have noted that organisations face genuine ambiguity about whether retaining a duplicate image of a person constitutes a separate data-holding obligation. The Personal Data Protection Commission has signalled interest in updated guidance on AI and data quality, but no formal consultation on image duplication has been announced as of July 2026.

The third and most commercially sensitive decision is who pays for the clean-up. Retroactive deduplication of large archives is not cheap. Cloud processing costs for scanning and reconciling a one-million-image catalogue using current AI-assisted tools run in the range of several thousand Singapore dollars per processing cycle, depending on resolution and database architecture. For smaller businesses — the hawker-centre vendors and neighbourhood retailers who have migrated to digital platforms under programs like SMEs Go Digital — that cost is prohibitive without subsidy or shared infrastructure.

The IMDA's SMEs Go Digital programme, which has supported more than 80,000 businesses since its launch, could be the vehicle for subsidised deduplication tooling. Whether the next round of approved solutions under the programme includes image-management and deduplication platforms will be a concrete indicator of how seriously the government is treating the problem. Decisions on that approved-vendor list are typically reviewed on an annual cycle, putting the next meaningful window at early 2027. Organisations that wait for that cycle risk compounding their archives further — and inheriting a more expensive problem when they finally act.