Abstract: Urban waterfront greenway is an important part of the construction of urban green space, which has multiple functions such as ecology, recreation, economy, and connection. As an important part of the urban landscape, the visual quality of the urban waterfront has a positive significance for shaping urban characteristics and style. Taking the riverside greenway on both sides of the Jiajiang River in Nanjing as an example, OpenStreetMap is used to obtain road network data to generate street view sampling points, and street view images are crawled with the help of Python and Baidu Street View API. The semantic segmentation of the image was carried out by deep learning technology, and the influencing mechanism of landscape elements and visual quality was analyzed in combination with the evaluation of scenic beauty. The results show that there is still room for improvement in the visual quality of the Nanjing waterfront greenway landscape, and the visual quality of the riverside greenway south of the Jiajiang River is better than that of the north of the Jiajiang River. Road Width Index, Green Visual Index, and Color Richness Index positively affect visual quality, while Building Visibility is restrained. The visual quality of urban waterfront greenway landscape is the result of the two-way effect of public visual perception and environmental space, and it is recommended to pay attention to the significant impact indicators and set priorities according to their impact intensity to effectively improve the landscape level of the waterfront greenway. This study reveals the factors that influence the visual perception of urban waterfront greenway landscapes and provides a reference for the optimization of waterfront greenway landscapes. 

 Keywords: deep learning; street view imagery; waterfront greenway; visual quality; landscape evaluation