Training deep convolution network with synthetic data for architectural morphological prototype classification

Abstract The use of architectural morphological analysis and generative design is an important strategy to interpret current designs and to propose novel ones. Conventional morphological features are defined based on qualitative descriptions or manually selected indicators, which include subjective bias, thus limiting generalizability. The lack of public architectural morphological datasets also leads to setbacks in data-driven morphological analysis. This study proposed a new method for generating topology-based synthetic data via a rule-based system and for encoding morphological information to promote morphological classification via deep learning. A deep convolution network, LeNet, which was modified in the output layer, was trained with synthetic data, including five spatial prototypes (central, linear, radial, cluster, and grid). The performance of the proposed method was validated on 40 practical architectural layouts. Compared to the ground truth, the proposed method provided an encouraging accuracy of 97.5% (39/40). Interestingly, the most possible mistakes of the LeNet were also understandable according to the architect's intuitive perception. The proposed method considered the statistical and overall characteristics of the training samples. This work demonstrated the feasibility and effectiveness of the deep learning network trained with synthetic architectural patterns for morphological classification in practical architectural layouts. The findings of this work could serve as a basis for further morpho-topology studies and other social, building energy, and building structure studies related to spatial morphology.