<b>​A paper "</b><b><i>Probing Urban Congestion Risks Through the Physical Environment: An Explainable Framework Based on Street View Imagery</i></b><b>"</b>

This project includes the code for the paper "Probing Urban Congestion Risks Through the Physical Environment: An Explainable Framework Based on Street View Imagery." It consists of 5 Python projects and 2 additional files:1. <b><i>SRF_embedding.py</i></b> - Constructs a spatial network of features formed through panoramic segmentation and edge expansion of images, utilizing GCN to extract road spatial feature (SRF) of the urban physical environment from street view images.2. <b><i>BEF_embedding.py</i></b> - Employs Grad-CAM to extract landscape features related to traffic tasks from street view images, which are then organized into a graph structure. Likewise, GCN is used to derive built environment feature (BEF) from street view imagery.3. <b><i>CAM_resnet.py</i></b> - An extended ResNet model enhanced with CAM that is capable of extracting hotspot areas in street view images based on the requirements of downstream tasks.4. <b><i>Congestion_risk_task_Concatenate.py</i></b> - Designed for predicting congestion risks by integrating RSF and BEF based on a feature concatenation strategy.5. <b><i>Congestion_risk_task_Fusion.py</i></b> - Also used for predicting congestion risks, this script integrates RSF and BEF based on a feature fusion strategy.6. <b><i>requirements.txt</i></b> - Lists all the necessary environments for the code to function properly.7. <b><i>read.txt</i></b> - Provides an overview of all the files included in this project.本项目包含论文《基于物理环境探究城市拥堵风险：一种基于街景影像的可解释框架》的配套代码，共包含5个Python项目与2个附加文件： 1. **`SRF_embedding.py`**：构建基于全景图像分割与边缘扩张生成的特征空间网络，借助图卷积网络（Graph Convolutional Network, GCN）从街景影像中提取城市物理环境的道路空间特征（Spatial Road Feature, SRF）。 2. **`BEF_embedding.py`**：采用梯度加权类激活映射（Gradient-weighted Class Activation Mapping, Grad-CAM）从街景影像中提取与交通任务相关的景观特征，并将其组织为图结构；同样借助图卷积网络从街景影像中提取建成环境特征（Built Environment Feature, BEF）。 3. **`CAM_resnet.py`**：一种集成类激活映射（Class Activation Mapping, CAM）的改进残差网络（ResNet）模型，可根据下游任务需求从街景影像中提取热点区域。 4. **`Congestion_risk_task_Concatenate.py`**：基于特征拼接策略整合道路空间特征与建成环境特征，用于预测城市拥堵风险。 5. **`Congestion_risk_task_Fusion.py`**：同样用于拥堵风险预测，该脚本基于特征融合策略整合道路空间特征与建成环境特征。 6. **`requirements.txt`**：列出代码运行所需的全部依赖环境。 7. **`read.txt`**：对本项目包含的所有文件进行概述说明。