Regularization of Building Boundaries in Satellite Images Using Adversarial and Regularized Losses

  title={Regularization of Building Boundaries in Satellite Images Using Adversarial and Regularized Losses},
  author={Stefano Zorzi and Friedrich Fraundorfer},
  journal={IGARSS 2019 - 2019 IEEE International Geoscience and Remote Sensing Symposium},
  • Stefano ZorziF. Fraundorfer
  • Published 1 July 2019
  • Computer Science
  • IGARSS 2019 - 2019 IEEE International Geoscience and Remote Sensing Symposium
In this paper we present a method for building boundary refinement and regularization in satellite images using a fully convolutional neural network trained with a combination of adversarial and regularized losses. Compared to a pure Mask R-CNN model, the overall algorithm can achieve equivalent performance in terms of accuracy and completeness. However, unlike Mask R-CNN that produces irregular footprints, our framework generates regularized and visually pleasing building boundaries which are… 

Figures and Tables from this paper

Machine-learned Regularization and Polygonization of Building Segmentation Masks

Experiments demonstrate that the proposed machine learning based approach for automatic regularization and polygonization of building segmentation masks is not only capable of obtaining accurate results, but also of producing visually pleasing building outlines parameterized as polygons.

Regularized Building Segmentation by Frame Field Learning

We add a frame field output to an image segmentation neural network to improve segmentation quality and provide structural information for a subsequent polygonization step. A frame field encodes two

Unconstrained Aerial Scene Recognition with Deep Neural Networks and a New Dataset

A large-scale dataset, called Mul-tiScene dataset, composed of 100,000 unconstrained images each with multiple labels from 36 different scenes, which allows two branches of studies: 1) developing novel CNNs for multi-scene recognition and 2) learning with noisy labels.

Map-Repair: Deep Cadastre Maps Alignment and Temporal Inconsistencies Fix in Satellite Images

An end-to-end deep learning approach is proposed which is able to solve inconsistencies between the input intensity image and the available building footprints by correcting label noises and, at the same time, misalignments if needed.

Polygonal Building Extraction by Frame Field Learning

A frame field output is added to a deep segmentation model for extracting buildings from remote sensing images and a polygonization algorithm is introduced that utilizes the frame field along with the raster segmentation.

Polygonal Building Segmentation by Frame Field Learning

A deep neural network is trained, which aligns a predicted frame field to ground truth contour data and a new polygonization algorithm is introduced, guided by the frame field corresponding to the raster segmentation, which produces more regular segmentations.

MultiScene: A Large-Scale Dataset and Benchmark for Multiscene Recognition in Single Aerial Images

A large-scale dataset, called MultiScene, composed of 100 000 unconstrained high-resolution aerial images is created, which can develop and evaluate deep networks for multiscene recognition using clean data and provide crowdsourced annotations of all images for the purpose of studying network learning with noisy labels.

PolyWorld: Polygonal Building Extraction with Graph Neural Networks in Satellite Images

PolyWorld is introduced, a neural network that directly extracts building vertices from an image and connects them correctly to create precise polygons and achieves not only notable quantitative results, but also produces visually pleasing building polygons.

ME-Net: A Multi-Scale Erosion Network for Crisp Building Edge Detection from Very High Resolution Remote Sensing Imagery

The proposed ME-Net is an effective and practical approach for detecting crisp building edges from VHR remote sensing imagery by embedding an erosion module (EM) in the network to crisp the edge and adding the Dice coefficient and local cross entropy of edge neighbors into the loss function to increase its sensitivity to the receptive field.



On Regularized Losses for Weakly-supervised CNN Segmentation

This approach simplifies weakly-supervised training by avoiding extra MRF/CRF inference steps or layers explicitly generating full masks, while improving both the quality and efficiency of training.

Image-to-Image Translation with Conditional Adversarial Networks

Conditional adversarial networks are investigated as a general-purpose solution to image-to-image translation problems and it is demonstrated that this approach is effective at synthesizing photos from label maps, reconstructing objects from edge maps, and colorizing images, among other tasks.

Normalized Cut Loss for Weakly-Supervised CNN Segmentation

This work proposes a new principled loss function evaluating network output with criteria standard in "shallow" segmentation, e.g. normalized cut which evaluates only seeds where labels are known while normalized cut softly evaluates consistency of all pixels.

Generative Adversarial Nets

We propose a new framework for estimating generative models via an adversarial process, in which we simultaneously train two models: a generative model G that captures the data distribution, and a

Building Extraction from Satellite Images Using Mask R-CNN with Building Boundary Regularization

This work presents a method combining Mask R-CNN with building boundary regularization, which produces better regularized polygons which are beneficial in many applications.

Mask R-CNN

This work presents a conceptually simple, flexible, and general framework for object instance segmentation that outperforms all existing, single-model entries on every task, including the COCO 2016 challenge winners.