sideseeing-network

Two connected research tracks on walkable urban space in São Paulo and Chicago.

1. Urban similarity model (analysis/). Describes São Paulo's 96 districts (reference: Brás, district 10) and Chicago's 77 Community Areas with street-morphology (M), built-form (B) and urban-function (U) feature families, and compares places descriptively. The Chicago-only and São Paulo–Chicago models were fitted on 6 October 2026.
2. Sidewalk network (network/). Field collection with a phone on a vest (video, GPS, motion sensors, voice), and the passage graph built from those recordings without a map. It aims at two papers: a West Loop + Brás dataset, and the passage graph.

Agents: read AGENTS.md first.

Start here

I want to…	Open
Know where the similarity model stands	analysis/docs/STATUS.md
Know where the sidewalk network stands	network/docs/STATUS.md
Read the cross-city model and its conclusions	model report
Explore the cross-city methods interactively	Cross-City Urban Explorer: python3 -m http.server 8765 --directory analysis/viz
Read the passage-graph work in plain English	GRAPH_WORK_EXPLAINED
Find every document of a track	analysis/README.md · network/docs/README.md
Find a dataset	Model inputs: analysis/data/README.md · field captures: network/data/README.md and Drive/RustFS inventory
Run or find a script	analysis/scripts/INDEX.md · network/README.md

Layout

README.md  AGENTS.md      this page; the guide for agents (CLAUDE.md points to it)
analysis/                 track 1: urban similarity model
  docs/                   status, protocol, decisions, per-city and cross-city reports, archive/
  viz/                    Cross-City Urban Explorer, a static site (deploy root)
  scripts/ config/ tests/ code, frozen method contracts, unit tests
  results/                published releases: SP/, Chicago/, SP_CHI/ (tracked)
  curio_nodes/            nodes and dataflows installed in Curio
  data/ cache/ work/      model inputs and intermediates (local, git-ignored)
network/                  track 2: sidewalk network
  docs/                   status, plans, protocols, phase notes, datasets/, archive/
  graph/                  the passage-graph package (python -m graph.<phase>)
  scripts/                route planning, transcription, Drive mount, transfers
  notebooks/              field-day analysis notebooks (local, not committed yet)
  data/                   captures and everything derived from them (local, git-ignored)

Development

Python 3.12. Two environments:

• .venv (this repo) for analysis/ and the route-planning scripts. pyproject.toml declares its dependencies; analysis/requirements.txt pins the tested ones.
• ../sideseeing-tools/venv (a sibling repo) for network/graph, the notebooks and anything needing torch, Whisper or SAM3. Heavy runs happen on desktop-01.

.venv/bin/python -m unittest discover -s analysis/tests        # analysis tests
node analysis/tests/check_viz_model.js                         # explorer vs published results
cd network && ../../sideseeing-tools/venv/bin/python -m pytest -q graph/tests   # graph tests