Development and validation of automated microscale walkability audit method

Area of Science:

• Urban planning
• Geographic Information Systems (GIS)
• Computer Vision

Background:

• Microscale walkability assessment is crucial for urban planning but traditionally labor-intensive and costly.
• Virtual audits using street view images offer a cost reduction but remain expensive for large-scale application.
• Existing computer vision models struggle to capture the full complexity of microscale walkability features.

Purpose of the Study:

• To develop and validate an automated system for microscale walkability audits.
• To combine custom-trained computer vision models, GIS, and street view imagery for comprehensive auditing.
• To assess the reliability of the automated audit against virtual audit results.

Main Methods:

• Utilized custom-trained computer vision models to analyze street view images.
• Integrated geographic information systems (GIS) for spatial data processing.
• Developed an automated audit process for a validated microscale walkability tool.
• Validated automated audit results against manual virtual audit data.

Main Results:

• The automated audit system successfully captured a complete set of validated microscale walkability items.
• Automated audit results demonstrated high reliability when compared to virtual audit findings.
• The developed method offers a significant improvement in scalability and cost-effectiveness.

Conclusions:

• Automated microscale walkability auditing using computer vision and GIS is a reliable and scalable approach.
• This automated method can effectively replace expensive virtual audits for large-scale assessments.
• The study paves the way for more efficient and comprehensive urban walkability evaluations.