Singapore's Digital Archives Get a Cleaner Picture: What Happened This Week on Duplicate Image Replacement

Singapore's push to clean up its sprawling digital records took a concrete step forward this week, as the Infocomm Media Development Authority confirmed it has expanded its AI-assisted duplicate image detection programme to cover a broader set of government data repositories. The move addresses a problem that archivists and digital records managers have quietly flagged for years: identical or near-identical images stored under different filenames, across multiple systems, consuming storage, skewing analytics and complicating search results for both officials and the public.

The timing is deliberate. Singapore's ongoing Smart Nation 2.0 agenda, which the Government has tied to a broader ambition of making the city-state a trusted AI hub in Southeast Asia, depends on data quality at every layer. Duplicate imagery sitting inside public databases is not an abstract annoyance — it distorts training data for machine-learning models, inflates cloud-storage costs and undermines the reliability of image-recognition tools being rolled out across agencies from the Housing Development Board to the Land Transport Authority.

What the Clean-Up Looks Like on the Ground

The practical scope of this week's developments stretches across several institutions. The National Library Board's digital collection at its Victoria Street headquarters has been processing a backlog of scanned archival photographs, many digitised from physical negatives held at the National Archives of Singapore on Canning Rise. Staff there have been using a combination of perceptual hashing — a technique that detects visually similar images even when file formats or resolutions differ — and manual review queues to flag duplicates before deletion or consolidation.

At Mediapolis in one-north, at least two media-tech firms working under IMDA's accelerator programmes have been piloting commercial duplicate-image-replacement pipelines for clients in the publishing and real estate sectors. Real estate listings platforms operating in Singapore have a particular exposure here: property portals routinely accumulate thousands of near-identical interior shots submitted by agencies, which pollute recommendation algorithms and slow page loads for prospective buyers already navigating a market where average HDB resale flat prices in mature estates like Bishan and Queenstown have stayed well above S$600,000 through the first half of 2026.

The challenge is not trivial in scale. A 2025 study by the Singapore Management University's School of Computing and Information Systems — published in January of that year — found that in a sample of publicly accessible government image repositories, roughly 18 percent of stored images had at least one near-duplicate within the same database. Extrapolated across the full scope of public digital holdings, that figure represents a significant and unnecessary storage and maintenance burden for agencies already managing tight IT procurement budgets.

Why Replacement, Not Just Deletion, Matters

The subtler policy point is the distinction between removing duplicates and replacing them with canonical, properly tagged master copies. Deletion alone can break hyperlinks embedded in older government web pages and e-services, particularly those referencing heritage images or public-education campaign materials. The recommended workflow now being formalised in the IMDA's updated Digital Government Blueprint guidelines involves identifying the highest-quality version of any duplicate cluster, designating it as the master record, and propagating that reference across all linked systems — rather than simply wiping the extras.

For individual Singaporeans, the most visible near-term impact will likely be on government e-service portals. MyInfo-linked applications, for example, which use uploaded identity photographs, have already seen improved processing times after internal duplicate-detection tools were applied to staging environments earlier this year. Citizens uploading documents through SingPass-connected services should expect fewer rejection flags caused by systems mistaking a resubmitted image for a duplicate when the two files differ only by compression level.

Agencies have been asked to complete an initial audit of their highest-traffic image repositories by the end of the third quarter of 2026. For organisations outside the public sector, IMDA has signalled it will publish updated technical guidelines for commercial platforms by August, giving firms operating everything from e-commerce catalogues to news archives a clearer standard to benchmark against as Singapore presses ahead with its data-quality agenda.