Artwalk

An AI Enhanced Urban Art Discovery App

Resumo

Artwalk is a mobile application designed to make urban street art discoverable, accessible, and understandable through the use of a distributed, AI-driven architecture. The app integrates four specialised agents—Camera Crew, Info Crew, Audio Crew, and Route Planner. These collectively analyse user-submitted photos, enrich them with contextual metadata, convert descriptions into audio narratives, and guide users through curated routes. Additionally, users can explore an interactive map of artworks currently limited to Munich, save both scanned and discovered works, and revisit them through a dedicated ‘Saved’ view.

Technologically, the system is built with React Native (frontend) and FastAPI (backend), using containerised services managed via Docker Compose. Local language models (LLaVA and LLaMA3), running via Ollama, support image classification and text generation. A custom text-to-speech (TTS) pipeline delivers podcast-style audio guides to improve accessibility, especially for visually impaired or audio-oriented users.

Artwalk draws from open data provided by streetartcities.com, and while the current scope is geographically limited, its modular design allows for scalable expansion. This project showcases how AI agents can be combined to enrich cultural exploration in public spaces, turning everyday environments into interactive, inclusive galleries.