CEU eTD Collection (2020); Taoufiq, Salma: Urban Buildings Classification Using CNN-based Hierarchical Models

CEU Electronic Theses and Dissertations, 2020
Author Taoufiq, Salma
Title Urban Buildings Classification Using CNN-based Hierarchical Models
Summary Urban building characterization is a complex problem with many involved parts whose solution can benefit the development of smart autonomous driving systems, the digital archiving of cultural artifacts, as well as the automation of real estate valuation. To contribute to this research area, this work focuses on a specific part of the problem: the classification of urban buildings from photographs of their facades into 10 categories: Church, Mosque, Synagogue, Buddhist Temple, House, Apartment Building, Mall, Store, Restaurant, and Office Building. For this purpose, a novel hierarchical multi-label CNNbased model is proposed. Based on the coarse-to-fine paradigm, a label tree is set up, and the model provides outputs corresponding to each level of the resulting hierarchy. Feedback from the coarser level to the finer one runs through the model using simple probabilistic notions encoded through a multiplicative layer connecting the parent coarse branch of the model to the child fine branch. The resulting model solves the urban building classification task while performing better than both a classical convolutional neural network, as well as an existing hierarchical model known as Branch Convolutional Neural Network (B-CNN), while using less parameters than its B-CNN counterpart.
Supervisor Benedek, Csaba; Nagy, Balázs
Department Mathematics MSc
Full texthttps://www.etd.ceu.edu/2020/taoufiq_salma.pdf

Visit the CEU Library.

© 2007-2021, Central European University