Junqiu - Momo

[twinscity]

06 Model

6.2 Site View

01 Render all

02 Render Part

03 Different Views

[twinscity]

06 Model

6.1 Project

01 Law map in site

Treat site as a rectangular: 90x280m

Compress the law map to this size, using the center point for positioning.

Put 20x20 cells in

02 Five Dimensions of a Database

Law: Plan

War: Structural components

Art: Elevation + Envelope

Myth: Courtyard + Enclosure

Language: Material

16 Materials Every Architect Needs to Know (And Where to Learn About Them) | ArchDaily

03 Model test (01 frame)

Human perspective

04 30-frame model generation

Blender all

Law

War

Myth

Art

Language Material

[twinscity]

05 Site

Site Analysis

[twinscity]

04 Street View

4.5 Match with All Map

Match the map once per second

Video Link: street

[twinscity]

04 Street View

4.4 Match with Language, Myth, and Art

01 YOLO Video Matching Logic

Combine the data from each SOM cell into a single matching text, then use the query corresponding to the YOLO class to perform the match.

Filter the results using the YOLO data from each frame:

score =

0.60 × semantic similarity

+ 0.20 × base strength

+ operation bonus

+ visual position score

+ diversity bonus

- usage penalty

matched_ALL_modal_NONEMPTY_aggregated_300dpi

[twinscity]

04 Street View

4.3 Match with War

01 Street skeletons

02 Match

Street Skeletons Data (geometric vector) match War Skeletons Data (geometric vector) and choose War cells

War Map Link: Prepared War SOM Matching with Pose Features

[twinscity]

04 Street View

4.2 Match with Law

01 Match Rules

Traffic sign confidence × box area --- Rule intensity --- Match to the Law cells

02 Whole Video Match Results

[twinscity]

04 Street View

4.1 Video Analysis

1.Original video

We downloaded a 30-second video of a Rotterdam street.

2.Yolo detection

Video Link: street

3.Classification with dataset

[twinscity]

03 Databases and Analysis

3.6 Cross-modal matching

01 Matching logic

Language provides semantic relationships, Theology provides narrative roles, and Art provides visual symbols.

three_modal_art_language_theology_matching.csv

02 Art, Language, Myth SOM

three_modal_compact_preview_stage_raw.png

[twinscity]

03 Databases and Analysis

3.4 Language

01 The Play in Language

Huizinga believed that the ‘play’ inherent in language is manifested in the way it creates a poetic world; I therefore chose poetry as my language corpus and crawled 1,000 poems from PoetryDB.

poetry_dataset.csv

I used the large language model Ollama to extract metaphors from the poems, identifying the subject and object of each metaphor, and utilised ‘subject / target / metaphor’ as the semantic input for subsequent SOM and cross-modal matching.

poetry_metaphors_1000_ollama.csv

3.5 Myth

01 The Play in Myth

homo_ludens_myth_index.csv

Huizinga argues that the fun of ‘Play’ in ‘Myth’ is embodied in the personification of myth; by extracting ‘personification gods’ and their narrative roles from mythological texts, these are utilised as a mythological narrative layer in subsequent spatial generation.

Six sources of mythological texts:

1. Rig Veda 10.90: Purusha Sukta

2. Poetic Edda: Völuspá

3. Hesiod: Theogony

4. Hesiod: Theogony — Theoi version

5. Bulfinch’s Mythology: The Age of Fable

6. The Remains of Hesiod the Ascraean

02 Myth Narrative Database

Using keywords relating to the personification of gods in Humoludens, 1,614 related texts were identified; after filtering, 330 narrative mythological texts were retained.

myth_related_paragraphs_strict.csv

03: Use Ollama to extract narrative roles

God role cards:

personified gods

mythological characters

narrative actions

social roles

myth_personified_god_role_cards.csv

[twinscity]

03 Databases and Analysis

3.3 Art

01 The Play in Art

Huizinga believed that there exists an instinct for play in art and painting, which also embodies religious sacredness; it is precisely this ‘play-function’—combined with elements of decoration, creation and mythology—that constitutes the ‘play’ within art.

Religious paintings

Consequently, I selected Christian paintings as the database for ‘Art’. Christian paintings contain numerous symbolic elements, which I have decomposed using YOLO. The database comprises 600 Christian paintings, segmented into 1,038 fragments, corresponding to the three-dimensional YOLO labels.

02 YOLO+ SAM

Dimension 1: Ritual

Angel, Chalice, Cross, Halo, Holy_Nails, Madonna

Dimension 2: Ornament

Decoration, Fleur_de_lis, Scroll

Dimension 3: Agon

Flag, Shield, Staff, Sword

03 Art SOM

The labels of the fragments of paintings are first translated into geometric prototypes, and then different spatial arrangements and graphic operations are generated based on Huizinga’s three dimensions of play.

In Huizinga’s theory, Language, Myth and Art are interconnected; they all generate symbols, order and cultural meaning through play. I therefore proceed to cross-modal matching of these three datasets.

[twinscity]

03 Databases and Analysis

3.2 War

01 100 War News Videos

War corresponds to Huizinga’s final characteristic of play: Social Group Formation. Therefore, the war video is not treated simply as a record of combat or violence, but as a process through which a separated community is produced.

02 Yolo Model to Detect Pose

It focuses on how human bodies appear, move, gather, disappear, or become hidden within the battlefield image.

Yolo Detection Video Link: War

03 SOM

These pose embeddings were used to train a SOM (Self Organize Map), so similar bodily actions were organized into nearby positions. Each sample then received a som_x / som_y coordinate.

04 Visualization

1.Create a conversion rule table

2.Map

[twinscity]

03 Databases and Analysis

3.1 LAW

01 Huizinga’s theory

Huizinga describes legal procedure as a rule-bound and ritualized form of play. Law corresponds to Huizinga’s final characteristic of play: Fixed Rules and Order.

02 Workflow

11 videos

Blip (describe videos)

Whisper (transcribe video transcripts)

Pyannote (Identify different speakers)

Whisper + Pyannote = speaker's identity

Merged (speaker's identity + on-screen descriptions = video content)

03 Analyze Court Rules

Rule Map Link: Markov Rule Topology: Final Edition

04 Visualize the Map

1.Construct a rule conversion table

2.Draw Map

Court Rules Video Link: law

[twinscity]

02 Theoretical Framework

Civilization at Play

Theory

Inspired by Huizinga’s Homo Ludens, this project explores play as a cultural force that precedes and shapes civilization. It investigates five databases—art, myth, language, war and law—to reveal how different social structures already contain rules, roles, conflicts, rituals and pleasures of play.

While engaging Constant Nieuwenhuys’ manifesto in New Babylon to challenge the modern city as a system that regulates behavior and extracts productivity.

The characteristics of play

Free Activity

Outside Ordinary Life

Bounded by Time and Space

Fixed Rules and Order

Social Group Formation

[twinscity]

01 Project Question

As a typical example of post-war modernist urban planning, Rotterdam’s reconstruction was based on functional zoning and strong rational order. By focusing on efficiency and solving urgent problems, this planning logic helped the city survive after the war, but it also gradually removed Rotterdam’s diversity and weakened its historical memory.

[twinscity]

Model

Law+ War

War

Connect the first and second floors.

Vertical circulation or an atrium connecting two levels.

Slope

[twinscity]

Model Refinement