Physical disorder of streets in 264 Chinese cities (2015)

Published: 15 August 2022| Version 1 | DOI: 10.17632/d3d4h5bvss.1
Contributors:
Jingjia Chen,
Long Chen,
Yan Li,
Ying Long

Description

These compressed files contain data and Python code to replicate the paper “Measuring physical disorder in urban street spaces: A large-scale analysis using street view images and deep learning”. The data arise from a large-scale analysis of physical disorder in 264 major cities in China. The purpose is to use street view images and deep learning model to estimate the physical disorder score of each sampling point, street and city. MobileNetV3 code: This is a PyTorch implementation of MobileNetV3 architecture as described in the paper. The trained models are also included in the code for prediction. Please see “README.md” for detailed procedures. Training samples: The image samples used to train the 15 physical disorder factors are included and can be used to train other machine learning models. Virtual audit tool: This tool is used to label images in a given directory, developed with a single Python script with GUI. Labeled images can be moved or copied into sub-directories, that are named with the assigned labels. Please see “README.md” for detailed procedures. Physical disorder results: The folder contains the results of the estimation of multi-scale physical disorder. The three shapefiles “china_cities.shp”, “china_streets.shp”, and “china_svipoints.shp” record physical disorder scores for 264 cities, 769,407 streets and 1,219,238 sampling points, respectively. The attributes of the data are explained below and detailed in “Metadata.doc”: 1. china cities.shp The attributes of this shapefile include: unique id of each city (ID), Chinese name of 264 cities (name), English name of 264 cities (nameEng), spatial distribution pattern of street physical disorder (mode), including: (a) scattered; (b) diffused along the urban expansion direction; (c) linear concentrated along arterial roads, and physical disorder value of each city (pdvalue), physical disorder value for each factor at each city (Field names are acronyms for factor names, e.g. "AB" is an acronym for abandoned buildings). See "Metadata.doc" for a cross-reference to factor names and their acronyms. 2. china_streets.shp The attributes of this shapefile include: unique id of each street (street_id), Chinese name of 264 cities (name), English name of 264 cities (nameEng), physical disorder value of each street (pdvalue), and physical disorder value for each factor at each street (Field names are acronyms for factor names, e.g. "AB" is an acronym for abandoned buildings). See "Metadata.doc" for a cross-reference to factor names and their acronyms. 3. china_svipoints.shp The attributes of this shapefile include: unique id of each sampling point (point_id), unique id of the street where it is located (street_id), physical disorder value of each point (pdvalue) and physical disorder value for each factor at each point (Field names are acronyms for factor names, e.g. "AB" is an acronym for abandoned buildings). See "Metadata.doc" for a cross-reference to factor names and their acronyms.

Files

Institutions

Tsinghua University

Categories

Urban Planning

License