Abstract: In order to avoid the homogeneous renewal of traditional commercial blocks for sustainable development mode of blocks, this research acquires images taken by experiencers in seven traditional commercial blocks in the old downtown area of Beijing and, using the convolutional neural network (CNN) technology, extracts five types of core elements of building, street, food consumption, behavioral activity and signboard advertisement that the experiencers focus on from the aforesaid images. The research has the seven blocks divided into two attributes (ontological and derived attributes) in accordance with the characteristics of the aforesaid elements and the perceived content of experiencers. Based on the proportion of the five types of elements and the two attributes, the research analyzes the relationship between experiencer preference and block development mode. Research results show that: 1) Experiencers prefer the derived attribute, and the difference in the proportion of the two attributes can be improved through the integration of relevant factors; 2) the development mode of the seven blocks is divided into the mode with a higher proportion of the derived attribute and the mode with a similar proportion of the two attributes, and the coexistence of the two modes is conducive to the improvement of the overall vitality of blocks in the old downtown area; 3) the mode featuring the balanced proportion of the two attributes is more conducive to the sustainable development of traditional commercial blocks. By directly extracting the five types of experiencer perception elements mentioned above, the research effectively identifies the problems existing in the development mode of traditional commercial blocks, which is helpful for the revival of traditional commercial blocks and the selection of multiple development strategies.