Delhi's Duplicate Image Problem: How the Capital Stacks Up Against London, Seoul and São Paulo

Delhi's municipal digital records contain an estimated tens of thousands of duplicate photographs — images of the same pothole, the same encroachment, the same broken streetlight filed multiple times across overlapping complaint portals — and the agencies responsible for cleaning them up are under growing pressure to act. The Municipal Corporation of Delhi launched an internal audit of its citizen-grievance image database in early 2026, flagging duplicate image replacement as a formal data-quality priority for the financial year ending March 2027.

The timing matters. Delhi's three formerly separate municipal corporations were unified under a single MCD structure in 2022, and the merger left behind a patchwork of legacy databases that, four years on, still generate redundant uploads at scale. Every time a resident photographs a damaged road on Mehrauli-Badarpur Road and files the same complaint on both the 311-equivalent MCD app and the Chief Minister's grievance portal, two near-identical images enter separate systems with no automated reconciliation layer between them.

What Other Cities Have Built

The gap between Delhi's current position and what peer cities have accomplished is measurable. Seoul's Smart City Operations Centre, housed at Seoul City Hall in Jung-gu, deployed a perceptual-hashing deduplication system across its public infrastructure image database in 2023 and reported a reduction in storage overhead as part of its annual digital governance report. London's City Hall has integrated the capital's pothole and street-defect reporting through a single Transport for London and borough-council API layer, meaning duplicate images are flagged at the point of upload rather than after the fact. São Paulo's Prefeitura ran a pilot on its Cidade Linda programme that used open-source image-fingerprinting tools to cut redundant complaint photographs from its maintenance database.

Delhi's situation is structurally more complicated. Unlike Seoul or London, the city's civic image data sits across at least four distinct platforms: the MCD's own grievance app, the Delhi government's Samadhan portal, the Public Works Department's internal filing system, and ward-level WhatsApp groups that feed unofficial but widely used reporting channels. Chandni Chowk ward offices, for instance, receive complaint images through both formal and informal routes simultaneously, creating duplication that no single agency currently has the mandate or the tools to reconcile end to end.

Delhi's Emerging Response

The MCD's IT department has been in procurement discussions for a deduplication middleware layer since at least January 2026, according to tender documentation listed on the Central Public Procurement Portal. The proposed system would apply hash-based matching to incoming complaint images before they are logged, preventing identical or near-identical files from entering the database as separate records. The contract, valued in the tender at under ₹2 crore, has not yet been awarded as of this week.

Separately, the Delhi Urban Shelter Improvement Board has been running a smaller-scale exercise on its housing survey photographs taken across resettlement colonies in areas including Rohini and Dwarka. That exercise, begun in the second half of 2025, has been cited internally as a proof-of-concept for a broader city-wide rollout, though no formal announcement has been made.

The comparison with São Paulo is instructive on cost. The Prefeitura's open-source approach kept implementation costs low by using publicly available perceptual-hashing libraries rather than proprietary platforms. Delhi's procurement route appears to favour a vendor-supplied solution, which carries higher licensing overhead but potentially faster deployment against MCD's existing infrastructure.

For residents filing complaints near landmarks like Connaught Place or lodging pollution-related grievances tied to the Yamuna riverbank, the practical upshot is this: until the deduplication system is in place, the same image filed twice through different channels will continue to generate two separate case numbers, two separate response timelines, and two entries in whichever dataset someone eventually tries to analyse. Agencies, ward councillors and researchers working with MCD data are advised to treat image-linked complaint counts as unreliable for statistical purposes until the audit is complete and the new middleware, if procured, has been running for at least one full reporting cycle.