Exploring the Impact of Built Environment Attributes on Social Followings Using Social Media Data and Deep Learning
Streets are an important component of urban landscapes and reflect the image, quality of life, and vitality of public spaces. With the help of the Google Cityscapes urban dataset and the DeepLab-v3 deep learning model, we segmented panoramic images to obtain visual statistics, and analyzed the impac...
Saved in:
Published in | ISPRS international journal of geo-information Vol. 11; no. 6; p. 325 |
---|---|
Main Authors | , , , |
Format | Journal Article |
Language | English |
Published |
Basel
MDPI AG
01.06.2022
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | Streets are an important component of urban landscapes and reflect the image, quality of life, and vitality of public spaces. With the help of the Google Cityscapes urban dataset and the DeepLab-v3 deep learning model, we segmented panoramic images to obtain visual statistics, and analyzed the impact of built environment attributes on a restaurant’s popularity. The results show that restaurant reviews are affected by the density of traffic signs, flow of pedestrians, the bicycle slow-moving index, and variations in the terrain, among which the density of traffic signs has a significant negative correlation with the number of reviews. The most critical factor that affects ratings on restaurants’ food, indoor environment and service is pedestrian flow, followed by road walkability and bicycle slow-moving index, and then natural elements (sky openness, greening rate, and terrain), traffic-related factors (road network density and motor vehicle interference index), and artificial environment (such as the building rate), while people’s willingness to stay has a significant negative effect on ratings. The qualities of the built environment that affect per capita consumption include density of traffic signs, pedestrian flow, and degree of non-motorized design, where the density of traffic signs has the most significant effect. |
---|---|
ISSN: | 2220-9964 2220-9964 |
DOI: | 10.3390/ijgi11060325 |