São Paulo Officials Tackle Millions in Duplicate Digital Images Across City Archives

São Paulo's municipal and academic institutions are sitting on a sprawling, largely unaudited problem: millions of duplicate digital images clogging government servers, slowing down public databases and inflating storage costs that ultimately fall on taxpayers. The issue has been building for years, but pressure to address it is intensifying in 2026 as prefeitura departments push to consolidate digital infrastructure ahead of a broader smart-city modernisation drive tied to the city's 2025–2028 municipal technology plan.

Duplicate image replacement — the process of identifying redundant files across a database, selecting a canonical version and systematically replacing broken or repeated references with a single authoritative copy — sounds technical. In practice, it touches everything from the Secretaria Municipal de Educação's photo archives of school construction projects to the geospatial imagery databases maintained by Geosampa, the city's official geographic information platform on Rua São Bento. When the same satellite tile or infrastructure photograph exists in fourteen slightly different compressed formats across four different server directories, nobody is well served.

Why the Problem Has Become Urgent Now

Storage is not free. The Tribunal de Contas do Município de São Paulo, which audits city finances, has flagged redundant data management as a contributor to unnecessary operating expenditure in municipal IT in past annual reviews, though the full dollar figure attributable specifically to image duplication has not been publicly itemised. Independent specialists in digital asset management say the Brazilian public sector broadly stores data at per-gigabyte costs of between R$0,80 and R$2,40 per month depending on contract structure — and large image archives can run into tens of terabytes.

At the Universidade de São Paulo's campus in Butantã, researchers working within the Instituto de Matemática e Estatística have been developing automated deduplication pipelines using perceptual hashing and embedding-based similarity scoring. The approach compares image fingerprints rather than raw file data, catching near-duplicates — the same photo saved at two resolutions, or with different metadata tags — that byte-level comparison would miss. Faculty members in that group have presented findings at domestic conferences arguing the technique could reduce image storage requirements in large civic repositories by between 30 and 60 percent, though that range depends heavily on how the original archive was managed.

The Secretaria Municipal de Inovação e Tecnologia, headquartered near Viaduto do Chá in the historic Centro, is understood to be evaluating vendor proposals for a citywide digital asset management overhaul that would include deduplication as a core feature. The city has not announced a contract award as of 4 July 2026.

What the Specialists and Advocates Are Recommending

Three broad positions have emerged in technical and policy circles here. One camp argues the city should procure an enterprise digital asset management platform — solutions from international vendors already tested in Bogotá and Lisbon have been discussed at São Paulo technology forums — with deduplication built into ingestion workflows from day one. A second group, prominent among open-source advocates connected to the Centro de Estudos e Pesquisas de Direito Sanitário in Perdizes, favours building on existing open tools and keeping the logic in-house to avoid long-term vendor lock-in. A third position, heard most clearly from fiscal watchdogs, is that any investment in deduplication tooling must be tied to a measurable storage-cost reduction target published before the contract is signed, not after.

The practical stakes extend beyond server economics. Public image archives underpin transparency portals, urban planning records and journalistic access to government documentation. When the same construction site photograph exists under three different file names in the Portal da Transparência, it creates confusion about which version carries official status — a procedural headache that researchers and journalists filing freedom-of-information requests through the e-SIC system on Rua Líbero Badaró have reported encountering.

What happens next is likely to be decided in the third quarter of 2026, when the prefeitura's technology secretariat is expected to finalise its procurement calendar. Experts advising the process say any institution serious about tackling the problem should begin with a full image-inventory audit before selecting tooling — mapping what exists, where it lives and who owns it. Without that baseline, even the best deduplication engine is working blind.