How Ballarat's Public Image Archive Ended Up Full of Duplicates — and What Comes Next

Ballarat City Council's digital asset management system holds tens of thousands of photographs accumulated over more than two decades of digitisation projects, tourism campaigns and heritage documentation work. A significant portion of those images — estimated internally at more than a third of the total archive — are duplicates, near-duplicates or low-resolution replacements of higher-quality originals that were filed separately and never merged. The problem did not appear overnight.

The roots trace back to at least 2003, when the first coordinated push to digitise physical photographic collections began across regional Victorian councils. Ballarat was an early mover, partly because of the scale of its gold-era heritage holdings and the importance of organisations like Sovereign Hill to the city's tourism identity. At the time, scanning equipment, file-naming conventions and storage infrastructure were all handled by different teams with different mandates. The result was predictable: the same image of Sturt Street's historic streetscape might exist as a 72-DPI scan filed under one project name, a 300-DPI version uploaded during a later grants application, and a JPEG thumbnail exported for a website refresh in 2011.

Why It Became Unmanageable

Three separate digitisation funding rounds — delivered through programs administered by the Public Record Office Victoria — brought fresh batches of material into the system without a consistent protocol for checking what already existed. Each grant had its own deliverables and timelines. Staff working on the Ballarat Heritage Precincts project in the early 2010s were largely separate from those managing the Visit Ballarat promotional photography library, and the two collections were stored on different servers before being consolidated into a shared environment around 2018.

The consolidation itself created new duplication. Automated migration tools flagged files with identical names but missed files with different names that contained identical image data. A photograph taken at the Ballarat Botanical Gardens on Wendouree Parade, for instance, might appear three times under three different filenames, each tagged with different metadata, none of them marked as authoritative.

Sovereign Hill's own photographic holdings add another layer. The living museum on Bradshaw Street maintains an independent archive for its interpretive and marketing purposes, and images shared between Sovereign Hill and council systems over the years were not always tracked. The Ballarat International Foto Biennale, which has run since 2005 and uses the Art Gallery of Ballarat on Lydiard Street North as a key venue, has contributed to the regional image pool through partnerships, adding professionally produced work that sometimes duplicated existing documentary shots of the same locations.

The Cost of Inaction

Storage costs are the most tangible pressure. Commercial cloud storage rates for large uncompressed image files have risen considerably since 2020, and a bloated archive with no deduplication protocol wastes budget that regional organisations can rarely spare. Beyond cost, the practical problem is findability. When a staff member needs a usable, high-resolution image of the Camp Street precinct for a grant acquittal or a media release, a search that returns forty results — most of them inferior copies of two or three originals — adds hours to a routine task.

The Victorian Government's Regional Digital Capacity grants, which have supported infrastructure upgrades at several Ballarat-based cultural organisations since 2022, have helped fund the audit work now underway. That audit, covering collections held by Ballarat City Council's libraries and heritage unit, is expected to identify candidate images for either deletion or replacement with a designated master file before the end of the 2026 calendar year.

Organisations working through this process typically start with automated hash-matching to find exact duplicates, then move to perceptual hashing for near-matches, before any human review. For Ballarat's collections, the human review stage matters most: a photograph taken in 1978 on Lydiard Street and a near-identical frame from 1982 are not the same record, even if a machine treats them as duplicates. Getting that distinction right is what separates a useful archive from a smaller but equally chaotic one. The organisations involved say they expect to establish a single authoritative file standard for all retained images by mid-2027 — a modest deadline, given how long the problem has been building.