// Athens workshop · Urban Data · Board Game · ASCAAD 2024
Athens City Game
We used machine learning to read how graffiti shapes the way people perceive Athens, then turned the findings into a route-planning board game.
Overview
Athens is one of the most graffiti-covered cities in the world. Spray paint sits on top of ancient stone, and it changes how a street feels: lively to some people, unsafe to others. We wanted to measure that effect across a whole district, and then make it something a visitor could play with.
Working in a five-person team, I collected and classified street-view images, ran the analysis that links graffiti to perception, and designed the board game and its visuals. The research was published at ASCAAD 2024 and presented in Barcelona.
What we built
29,820 street-view images7,455 sampling points4 graffiti types6 perception scores1 board game · 1 paper
My contribution at a glance
Street-view data & ML classificationCorrelation analysisBoard game mechanicsMap, cards & packaging design
Question
01
Does graffiti change how a street feels?
Graffiti is part of Athens’ identity, but its effect on how people read a street is hard to pin down from a few walks. We looked at the first district street by street, asking two questions at once: what kind of graffiti is here, and how does this place feel?
A
What is on the wall
none, tags, throw-ups, or full works
B
How the street feels
beautiful, lively, clean, safe, depressing, wealthy
C
How they relate
mapped across the whole district
Method
02
Reading a city through street view
We sampled the street network of Athens’ first district every 20 metres and downloaded four street-view images at each point, one per direction. Two models then read every image: one sorted the graffiti into four types, and one scored the scene on six perception dimensions, trained on the Place Pulse dataset of crowd-sourced judgments.
Street networkpoints every 20 m
→
Street view4 headings per point
→
Graffiti modelResNet · 4 types
→
Perception modelPlace Pulse · 6 scores
→
Statistics & mapsANOVA · GWR
Collect
7,455 points29,820 street-view images across the first district.
Classify
4 graffiti typesNone, tag, throw-up, and work, labelled with a ResNet image classifier.
Compare
6 perception scoresTested across graffiti types and mapped with geographically weighted regression.
Findings
03
Graffiti makes streets livelier, and less comfortable
Regression and correlation tests showed that graffiti does not push perception in one direction. Its presence goes together with streets that feel both more lively and more depressing, while they read as less safe, clean, wealthy, and beautiful.
Rises with graffiti
Lively and depressing scores both go up: graffiti gives a street energy and, at the same time, a sense of neglect.
Falls with graffiti
Safe, clean, wealthy, and beautiful scores go down, and the strength of the effect varies from one part of the district to another.
Design
04
Turning the map into a game
A map of correlations is hard for a visitor to use. So we built Athens Polis, a pen-and-paper game played on the same map. Each spot on the board is a real graffiti location, and each area carries a safety score from the analysis. Players plan a graffiti-visiting route, trading off how many spots they can reach against how safe the streets in between feel, and learn the city by playing with its data.
Draw 2 target cardsgoals for the round
→
Flip an action cardnext spot to visit
→
Draw the routeconnect spots on the map
→
Repeat8 turns per game
→
Score the routespots + safety + areas
Map
Data as the boardGraffiti spots by type, and area scores from the regression analysis.
Cards
Targets, actions, rewardsPlus landmark cards that introduce the city along the way.
Players
1–3 playersAbout 15 to 25 minutes, with only pens and paper maps.
Publication
05
From workshop to conference
The research was accepted at ASCAAD 2024, the 12th International Conference of the Arab Society for Computation in Architecture, Art and Design, and presented in Barcelona in November 2024.
X. Guan, Y. Gao, J. Cai, X. Xu, and H. Tu. Athens-Polis: Exploring the Correlation between Urban Graffiti and Cognitive Indices in Athens Through Machine Learning ↗. ASCAAD 2024, Computational Design and Territorial Resilience.
Outcome
One dataset, two outputs: a paper and a game.
The same analysis produced a paper accepted at ASCAAD 2024 and a playable board game, which turns its correlations into rules a visitor can use: more spots, safer streets, wider coverage.