Duplicate Images in Adelaide's Digital Archives: What Happens Next and the Key Decisions Ahead

South Australia's major cultural and government institutions are sitting on a problem that has quietly compounded for years: vast digital image libraries riddled with duplicates, conflicting metadata and redundant files that consume storage budgets, slow public search tools and — in some cases — result in the wrong image being served to researchers, journalists and the general public. The question now is not whether to act, but how.

The issue has crystallised in 2026 as several Adelaide-based agencies undertake concurrent digitisation and migration projects. The State Library of South Australia on North Terrace, which holds one of the country's most significant photographic collections, is mid-way through a multi-year digitisation program. Meanwhile, History Trust of South Australia, based at the Migration Museum on Kintore Avenue, is consolidating collections that span physical and born-digital material acquired over more than three decades. When both institutions push assets into shared discovery platforms, duplicate images — sometimes dozens of versions of a single photograph at varying resolutions and with inconsistent rights metadata — follow them in.

Why the Stakes Are Higher Now

The timing matters. South Australia's Lot Fourteen precinct on North Terrace is home to a growing cluster of data, space and technology organisations, and the state government has positioned digital infrastructure as a core economic plank alongside AUKUS and the hydrogen jobs plan. Cultural institutions are under pressure to modernise their public-facing platforms and contribute usable, clean data to broader discovery ecosystems. Duplicate images are not a minor housekeeping issue in that context — they are a data-quality failure that undermines the credibility of everything built on top of the collection.

The practical consequences are tangible. A researcher requesting a high-resolution image of, say, the 1936 Centenary celebrations on King William Street may receive a low-resolution scan with incomplete copyright attribution simply because it was indexed first and the better file sits buried under a near-identical record created during a separate digitisation batch. For institutions licensing images commercially, that confusion has direct financial implications.

Internationally, the cultural sector has moved toward automated deduplication tools that use perceptual hashing — a technique that generates a fingerprint for each image based on visual content rather than file name or metadata — to flag near-identical files for human review. The Australian Institute for Machine Learning, headquartered at Lot Fourteen, has expertise directly applicable to this kind of large-scale image analysis, and discussions between technology partners and cultural institutions about potential collaboration are understood to be at an exploratory stage, though no formal agreements have been announced publicly.

The Decisions That Cannot Be Delayed

Three choices will define how this plays out. First, institutions must decide which file version to designate as the canonical record when duplicates are confirmed — highest resolution, earliest acquisition date, or most complete rights metadata. There is no universal standard, and South Australian agencies are not yet working from a shared policy framework on this question.

Second, they must determine what happens to the replaced files. Deletion carries risk: a duplicate may be the only surviving version with provenance information attached, even if the image itself is inferior. The State Records Act 2000 governs disposal decisions for government agencies, and cultural institutions operate under their own collection management policies, meaning the legal and procedural pathways differ between organisations even when they share a discovery platform.

Third — and most consequentially for the public — institutions must decide how transparent to be about the replacement process. When an image that has been publicly accessible under a specific URL or catalogue number is withdrawn or superseded, researchers who cited it in published work face broken references. The British Library and Libraries Australia have both grappled with persistent identifier policy in recent years, and South Australian institutions will need to land on a local position.

The State Library's digitisation program is scheduled to reach a major milestone by mid-2027, which sets a practical deadline. Institutions that wait until then to resolve their deduplication and replacement policies will be managing the problem retrospectively at scale, rather than building clean processes from the outset. The window to get ahead of it is narrowing.