Projects‎ > ‎

Architectural Style Classification using MLLR

Architectural style classification differs from standard classification tasks due to the rich inter-class relationships between different styles, such as re-interpretation, revival, and territoriality. In this paper, we adopt Deformable Part-based Models (DPM) to capture the morphological characteristics of basic architectural components and propose Multinomial Latent Logistic Regression (MLLR) that introduces the probabilistic analysis and tackles the multi-class problem in latent variable models. Due to the lack of publicly available datasets, we release a new large-scale architectural style dataset containing twenty-five classes. Experimentation on this dataset shows that MLLR in combination with standard global image features, obtains the best classification results. We also present interpretable probabilistic explanations for the results, such as the styles of individual buildings and a style relationship network, to illustrate inter-class relationships.


The 25-class dataset used in the ECCV paper can be assessed through the link below.

(The download link has been updated 16/09/2014).

downloadlink (google) or downloadlink (baidu)

The ten styles used in our first experiments are: (randomly 30 images used for training and 10-fold cross validation)
American craftsman style
Baroque architecture
Chicago school architecture
Colonial architecture
Georgian architecture
Gothic architecture
Greek Revival architecture
Queen Anne architecture
Romanesque architecture
Russian Revival architecture

For a better evaluation of inter-class relationships between architectural styles, we are further collecting a larger number of styles (for now 66 styles), and tries to manage more metadata such as the name and place of the buildings. Please keep in track if you're interested in the task.