Geographic information system navigation services are now incorporating psychological well‐being as a factor when devising navigation routes. However, challenges such as limited data, method generalizability, and subjective human perception remain unresolved. Therefore, a general humanized path‐navigation method that effectively quantifies human emotional perception demands is required. In this study, we designed a deep learning model using a large, crowdsourced dataset to predict emotional responses to street‐view images. Our method enhances urban path planning, thus providing comprehensive emotional benefits. Our approach and several goal‐oriented methods were applied in Nanjing, and the findings were compared via comparative analyses and questionnaire surveys. The results confirmed that our proposed method outperforms utilitarian goal‐driven path planning methods in terms of subjective perception. This study provides a widely available technique for high‐quality navigation planning that meets psychological perception needs and offers a valuable guidance for the research on humanized geographic information services.