An open retail boundary dataset for South Korea using open data and computer vision technique

Area of Science:

• Geographic Information Systems (GIS)
• Computer Vision
• Retail Analytics

Background:

• Understanding retail sector dynamics requires accurate delineation of retail boundaries.
• Existing research lacks specific studies on retail boundary delineation in the South Korean context.
• Accurate retail geography is crucial for policy-making and economic analysis.

Purpose of the Study:

• To propose and validate a novel methodology for delineating retail boundaries across South Korea.
• To generate a comprehensive dataset of retail boundaries for South Korea.
• To support research on retail gentrification, pandemic recovery, and e-commerce impacts.

Main Methods:

• Utilized diverse retailer and building datasets.
• Developed a computer vision-based framework incorporating a deep ensemble voting technique.
• Delineated 6,636 distinct retail boundaries.

Main Results:

• Successfully delineated 6,636 retail boundaries across South Korea.
• Validated the delineated boundaries against existing reference data.
• Created a foundational dataset for retail geography research.

Conclusions:

• The proposed framework effectively delineates retail boundaries in South Korea.
• The generated dataset enhances understanding of retail geography and its dynamics.
• This resource supports future research on retail resilience and urban development.