Abstract
In this paper we present an approach for semantic interpretation of facade images based on a Convolutional Network. Our network processes the input images in a fully convolutional way and generates pixel-wise predictions. We show that there is no need for large datasets to train the network when transfer learning is employed, i. e., a part of an already existing network is used and fine-tuned, and when the available data is augmented by using deformed patches of the images for training. The network is trained end-to-end with patches of the images and each patch is augmented independently. To undo the downsampling for the classification, we add deconvolutional layers to the network. Outputs of different layers of the network are combined to achieve more precise pixel-wise predictions. We demonstrate the potential of our network based on results for the eTRIMS (Korč and Förstner, 2009) dataset reduced to facades.
Highlights
Deep Learning and especially Convolutional Networks (ConvNets) are gaining more and more interest in recent years
Others are image classification, where a single label is assigned to a whole image (Krizhevsky et al, 2012, Simonyan and Zisserman, 2014, Szegedy et al, 2015) or semantic segmentation with Fully Convolutional Networks (Long et al, 2015)
With regard to both the growing need for semantic building models as well as the developing field of Deep Learning, we present in this paper an approach based on ConvNets to segment and classify facade-elements in images
Summary
Deep Learning and especially Convolutional Networks (ConvNets) are gaining more and more interest in recent years. Others search for repetitive patterns on two-dimensional (2D) input images to segment individual facades (Wendel et al, 2010) or try to find windows with the help of implicit shape models (Reznik and Mayer, 2008). With regard to both the growing need for semantic building models as well as the developing field of Deep Learning, we present in this paper an approach based on ConvNets to segment and classify facade-elements in images.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
