As an important part of urban environment construction, garden landscape design, its spatial layout and aesthetic effect directly affect people’s quality of life. This study explores the application of computer vision technology in garden landscape design, and quantitatively evaluates the spatial layout and aesthetic features of gardens through the methods of image semantic segmentation, spatial syntactic analysis and SD semantic difference method. Typical gardens in A and B were selected for the study, and parameters such as spatial integration degree and connection value were calculated using Depthmap software, and aesthetic evaluation was carried out through 160 valid questionnaires. The results show that the global integration degree of garden B is 0.8266, which is 0.1781 higher than that of garden A. The t-test p-value of the road integration degree of the two gardens is 0.0244, which indicates that there is a significant difference in accessibility; the spatial undulation and change factor scores the highest in the aesthetics evaluation (0.7496), and transcendence feeling scores the lowest (0.1525). The study shows that computer vision technology can effectively identify the spatial characteristics of the garden, spatial syntax parameters can objectively reflect the rationality of the layout, while the aesthetic evaluation needs to be comprehensively optimized at the three levels of physical, emotional, and intentional contexts.
