Graffiti Guard

Turn a street photo into a private graffiti review with highlighted regions, clear next steps, and a shareable reference on iPhone or iPad.

Detect likely graffiti in photos and camera captures without uploading images.

Graffiti Guard is designed for residents, field teams, councils, and maintenance crews that need a quick, private first review. Its bundled Core ML model runs entirely on your iPhone or iPad and highlights likely graffiti for human verification.

The iPad workflow is built for inspections: drop in a photo, adjust confidence, review highlighted regions, then share a referenced summary with the result, confidence, processing time, and recommended next step.

Graffiti Guard is designed to shorten the first-review and handoff steps. On-device processing avoids image uploads and inference-server dependency, while consistent references help teams route visible incidents. Actual time or cost savings depend on each organisation's workflow and should be measured in a pilot.

Why it exists

Recurring unwanted graffiti creates cleanup pressure and surface damage in Melbourne and cities worldwide, including Paris, Brussels, Bucharest, and Detroit. Graffiti Guard helps teams review visible incidents earlier and coordinate maintenance for cleaner, better-kept public spaces.

Key features:

• Private, on-device image processing
• No inference server, account, or internet connection required
• Camera, Photos, Files, and iPad drag-and-drop support
• Adjustable confidence threshold
• Shareable inspection summaries with clear next steps
• Universal iPhone and iPad interface

Created by Pierre-Henry Soria (PierreHenry.dev), an AI software engineer and consultant specialising in computer vision, on-device machine learning, privacy-conscious systems, and production AI deployment. Graffiti Guard is the mobile companion to his open-source Graffiti Detection AI Model Python library.

Graffiti Guard detects likely graffiti in images. It does not predict future vandalism and should not replace human review.