ABSTRACT:In this paper we present a pipeline for high quality semantic segmentation of building facades using Structured Random Forest (SRF), Region Proposal Network (RPN) based on a Convolutional Neural Network (CNN) as well as rectangular fitting optimization. Our main contribution is that we employ features created by the RPN as channels in the SRF. We empirically show that this is very effective especially for doors and windows. Our pipeline is evaluated on two datasets where we outperform current state-of-the-art methods. Additionally, we quantify the contribution of the RPN and the rectangular fitting optimization on the accuracy of the result.

No abstract

No abstract

ABSTRACT:In this paper we present a bottom-up approach for the semantic segmentation of building facades. Facades have a predefined topology, contain specific objects such as doors and windows and follow architectural rules. Our goal is to create homogeneous segments for facade objects. To this end, we have created a pixelwise labeling method using a Structured Random Forest. According to the evaluation of results for two datasets with the classifier we have achieved the above goal producing a nearly noise-free labeling image and perform on par or even slightly better than the classifier-only stages of state-of-the-art approaches. This is due to the encoding of the local topological structure of the facade objects in the Structured Random Forest. Additionally, we have employed an iterative optimization approach to select the best possible labeling.

No abstract