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

Zorzi, Stefano; Fraundorfer, Friedrich

doi:10.1109/igarss.2019.8900337

Cited by 23 publications

(26 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In the following work, Cheng et al [26] introduced a network based on a polar representation of active contours which prevent self-intersections and enforces outlines to be even closer to the ground truth. Work most related to ours is Zorzi et al [7], which looked at the problem differently. The authors of this paper trained the regularization network in an unsupervised manner using adversarial losses together with Potts [27,28] and normalized cut [28] regularization losses which embedded additional knowledge about building boundaries from the intensity image to the network.…”

Section: Related Workmentioning

confidence: 99%

“…The authors of this paper trained the regularization network in an unsupervised manner using adversarial losses together with Potts [27,28] and normalized cut [28] regularization losses which embedded additional knowledge about building boundaries from the intensity image to the network. In our work, we extend the algorithm proposed in [7] redefining the training procedure and the architecture of the regularization network to obtain better results both in qualitative and quantitative terms.…”

Section: Related Workmentioning

confidence: 99%

“…The interest in and the development of new methodologies was also motivated by the current existence of several public benchmark datasets, like INRIA [1], SpaceNet [2], and Crow-dAI [3]. The classical approaches in this research field mostly focused on the assignment of the semantic class to each pixel in the image, obtaining classification masks as output [4][5][6][7]. However, for many applications, the more advanced output in form of vector information is under demand.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Machine-learned Regularization and Polygonization of Building Segmentation Masks

Zorzi

Bittner

Fraundorfer

2021

2020 25th International Conference on Pattern Recognition (ICPR)

Self Cite

View full text Add to dashboard Cite

We propose a machine learning based approach for automatic regularization and polygonization of building segmentation masks. Taking an image as input, we first predict building segmentation maps exploiting generic fully convolutional network (FCN). A generative adversarial network (GAN) is then involved to perform a regularization of building boundaries to make them more realistic, i.e., having more rectilinear outlines which construct right angles if required. This is achieved through the interplay between the discriminator which gives a probability of input image being true and generator that learns from discriminator's response to create more realistic images. Finally, we train the backbone convolutional neural network (CNN) which is adapted to predict sparse outcomes corresponding to building corners out of regularized building segmentation results. Experiments on three building segmentation datasets demonstrate that the proposed method is not only capable of obtaining accurate results, but also of producing visually pleasing building outlines parameterized as polygons.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Machine-learned Regularization and Polygonization of Building Segmentation Masks

Zorzi

Bittner

Fraundorfer

2021

2020 25th International Conference on Pattern Recognition (ICPR)

Self Cite

View full text Add to dashboard Cite

show abstract

“…Buildings are one of the most significant elements in urban landscapes and are highly dynamic [1]. Automatic extraction of buildings is a long-standing problem [2][3][4][5][6][7] in urban scene classification, land use analysis, and automated map updating. The research related to building extractions can be broadly categorized as "building region detection" and "building edge detection".…”

Section: Introductionmentioning

confidence: 99%

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

Wen

Zhang

et al. 2021

Remote Sensing

View full text Add to dashboard Cite

The detection of building edges from very high resolution (VHR) remote sensing imagery is essential to various geo-related applications, including surveying and mapping, urban management, etc. Recently, the rapid development of deep convolutional neural networks (DCNNs) has achieved remarkable progress in edge detection; however, there has always been the problem of edge thickness due to the large receptive field of DCNNs. In this paper, we proposed a multi-scale erosion network (ME-Net) for building edge detection to crisp the building edge through two innovative approaches: (1) embedding an erosion module (EM) in the network to crisp the edge and (2) adding the Dice coefficient and local cross entropy of edge neighbors into the loss function to increase its sensitivity to the receptive field. In addition, a new metric, Ene, to measure the crispness of the predicted building edge was proposed. The experiment results show that ME-Net not only detects the clearest and crispest building edges, but also achieves the best OA of 98.75%, 95.00% and 95.51% on three building edge datasets, and exceeds other edge detection networks 3.17% and 0.44% at least in strict F1-score and Ene. In a word, the proposed ME-Net is an effective and practical approach for detecting crisp building edges from VHR remote sensing imagery.

show abstract

“…In the phase of data preparation, we note that although massive high-resolution aerial images can be effortlessly obtained from remote sensing data platforms, such as Google Earth, 1 it is extremely time-and labor-consuming to yield their corresponding multiple scene labels. To alleviate such annotation burden, in this article, we resort to crowdsourced data, e.g., OpenStreetMap 2 (OSM) annotations, which has been proven to be successful in generating image-level labels [27], [28], [35] and pixel-wise footprints [12], [36] for training deep networks. However, we observe that OSM data might suffer from two common defects, incompleteness and incorrectness, which could introduce severe noise into image labels.…”

mentioning

confidence: 99%

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

Hua

Mou²,

et al. 2022

IEEE Trans. Geosci. Remote Sensing

View full text Add to dashboard Cite

Aerial scene recognition is a fundamental research problem in interpreting high-resolution aerial imagery. Over the past few years, most studies focus on classifying an image into one scene category, while in real-world scenarios, it is more often that a single image contains multiple scenes. Therefore, in this article, we investigate a more practical yet underexplored task-multiscene recognition in single images. To this end, we create a large-scale dataset, called MultiScene, composed of 100 000 unconstrained high-resolution aerial images. Considering that manually labeling such images is extremely arduous, we resort to low-cost annotations from crowdsourcing platforms, e.g., OpenStreetMap (OSM). However, OSM data might suffer from incompleteness and incorrectness, which introduce noise into image labels. To address this issue, we visually inspect 14 000 images and correct their scene labels, yielding a subset of cleanly annotated images, named MultiScene-Clean. With it, we can develop and evaluate deep networks for multiscene recognition using clean data. Moreover, we provide crowdsourced annotations of all images for the purpose of studying network learning with noisy labels. We conduct experiments with extensive baseline models on both MultiScene-Clean and MultiScene to offer benchmarks for multiscene recognition in single images and learning from noisy labels for this task, respectively. To facilitate progress, we make our dataset and trained models available on https://gitlab.lrz.de/ai4eo/reasoning/multiscene.

show abstract

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

Cited by 23 publications

References 8 publications

Machine-learned Regularization and Polygonization of Building Segmentation Masks

Machine-learned Regularization and Polygonization of Building Segmentation Masks

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

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

Contact Info

Product

Resources

About