With recent development of analytical tools in the past decade, a series of approaches evaluating perceived street qualities have been developed. Nevertheless, street attractiveness as an important concern of contemporary urban design is still lacking quantitative measurements. As a response, we are attempting to develop an evaluation of the perceptual-based attractiveness of streets which usually depended on subjective experience. With multi-sourced urban data and machine learning algorithms, a human-centered evaluation has been developed to measure street attractiveness from three dimensions: visual quality, network accessibility and functional diversity. Specifically, street view images and machine learning algorithms were applied to quantify visual quality intelligently. Spatial design network analysis (sDNA) was used to measure street network accessibility. The entropy of points of interest (POIs) was used to assess functional diversity on streets. Beijing and Shanghai, two megacities from China were selected for case study. Analytic hierarchy process (AHP) was applied to integrate these three key dimensions to evaluate street attractiveness. Furthermore, the satisfactory accuracy of the approach has been verified by further validation. The analytical approach helps to quantify the degree of street attractiveness comprehensively, owing to the application of multi-source urban data. In short, this study contributes to the development of a human-centered and systematic measurement of street attractiveness across large-scale areas, which benefits planning practitioners to get information more efficiently and precisely. Findings achieved from this study would contribute to bring in a human-oriented perspective into morphometrics and newly developed analytical tools.