Jeong Byeonghwa, Song Juhwan, Kim Younghoon
R&D Associate, Human Geography Group, Oak Ridge National Laboratory, Oak Ridge, USA.
Ph.D. candidate, Department of Geography Education, Korea National University of Education, Cheongju, South Korea.
Sci Data. 2025 Apr 14;12(1):622. doi: 10.1038/s41597-025-04958-1.
Although delineating retail boundaries is important to explore and comprehend the dynamics of the retail sector, it is hard to find studies specifically addressing it in the South Korean context. This study fills this gap by proposing new retail boundaries across South Korea. To achieve this goal, we employed a variety of retailers and building datasets and proposed a unique computer vision-based framework with a deep ensemble voting technique. As a result, we delineated 6,636 distinct retail boundaries that were validated against existing reference retail boundaries. These newly delineated retail boundaries provide valuable insights for researchers, governments, and other relevant stakeholders by enhancing their understanding of retail geography. This dataset can be used as a foundational resource for analyses on topics such as pandemic recovery, retail gentrification, and the resilience of retail spaces in response to e-commerce growth, ultimately contributing to more robust retail sector research in South Korea.
虽然划定零售边界对于探索和理解零售行业的动态很重要，但在韩国背景下，很难找到专门针对此进行研究的文献。本研究通过提出韩国各地新的零售边界来填补这一空白。为实现这一目标，我们使用了各种零售商和建筑数据集，并提出了一个独特的基于计算机视觉的框架以及深度集成投票技术。结果，我们划定了6636个不同的零售边界，并与现有的参考零售边界进行了验证。这些新划定的零售边界通过增强对零售地理的理解，为研究人员、政府和其他相关利益者提供了有价值的见解。该数据集可作为基础资源，用于分析诸如疫情复苏、零售绅士化以及零售空间应对电子商务增长的韧性等主题，最终有助于韩国零售行业研究更加稳健。
