Quantifying the usage of small public spaces using deep convolutional neural network.

Area of Science:

• Urban Planning and Design
• Computer Vision
• Environmental Psychology

Background:

• Small public spaces are crucial for urban activities but lack efficient usage quantification methods.
• Existing approaches fail to effectively measure and analyze the utilization of these vital community assets.

Purpose of the Study:

• To develop and validate a novel deep convolutional neural network (CNN) approach for quantifying small public space usage.
• To bridge the literature gap by providing a reliable and robust method for analyzing space utilization through video data.

Main Methods:

• Deployed photographic devices to record videos of small public spaces.
• Utilized a deep convolutional neural network (CNN) to detect and track people within the recorded videos.
• Converted image-based positions to real-world projected coordinates for accurate spatial analysis.

Main Results:

• The developed CNN method successfully quantified the usage of small public spaces.
• Experiments in a Beijing residential community validated the approach's accuracy and robustness.
• Demonstrated effective and efficient measurement of space utilization.

Conclusions:

• The deep convolutional neural network (CNN) method provides an effective and efficient solution for quantifying small public space usage.
• This approach offers a reliable tool for urban planners and researchers to understand and optimize the use of public spaces.