Cartographic Relief Shading with Neural Networks

Jenny, Bernhard; Heitzler, Magnus; Singh, Deepesh; Farmakis-Serebryakova, Marianna; Liu, Jeffery Chieh; Hurni, Lorenz

doi:10.1109/tvcg.2020.3030456

Cited by 28 publications

(35 citation statements)

References 62 publications

(70 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Moreover, dropout is used at each downsampling and upsampling step with increasing dropout rates towards the bottleneck. It significantly prevents overfitting by avoiding the units co-adapting too much as well as enables to train and combine many different network architectures by randomly sampling a "thinned" network consisting of all the units that survive dropout (Srivastava et al, 2014;Jenny et al, 2020). The network is shown in Figure 5.…”

Section: Road Extraction With U-netmentioning

confidence: 99%

A Novel Data Augmentation Method to Enhance the Training Dataset for Road Extraction From Swiss Historical Maps

Jiao

Heitzler

Hurni

2022

ISPRS Ann. Photogramm. Remote Sens. Spatial Inf. Sci.

View full text Add to dashboard Cite

Abstract. Long-term retrospective road data are required for various analyses (e.g., investigation of urban sprawl, analysis of road network evolution). Yet, it is challenging to extract roads from scanned historical maps due to their dissatisfying quality. Although deep learning has been exerting its superiority in image segmentation, its application to road extraction from historical maps is rarely seen in existing studies. Deep learning usually requires quite large amounts of training data, which is time-consuming and tedious to label. Data augmentation can to some extent solve this issue. The existing data augmentation techniques vary each training sample as a whole (e.g., rotation, flipping). But some features or symbols on maps will never occur in practice when they are rotated or flipped (e.g., numbers, labels). To solve this problem and to further improve the diversity of training samples, we propose a novel data augmentation method, which varies the target features instead of the whole training sample. The method is validated by applying it to road extraction from the historical Swiss Siegfried map. The experiment results show the effectiveness of the proposed method.

show abstract

Section: Road Extraction With U-netmentioning

confidence: 99%

A Novel Data Augmentation Method to Enhance the Training Dataset for Road Extraction From Swiss Historical Maps

Jiao

Heitzler

Hurni

2022

ISPRS Ann. Photogramm. Remote Sens. Spatial Inf. Sci.

View full text Add to dashboard Cite

show abstract

“…Hence, a final cropping operation is applied to remove border pixels for which no meaningful results are expected. Finally, drop out layers are inserted to prevent overfitting, a technique that has been proven useful for similar applications [20]. The final convolutional layer makes use of a softmax activation function.…”

Section: Network Architecturementioning

confidence: 99%

Terrain Segmentation Using a U-Net for Improved Relief Shading

Farmakis-Serebryakova

Heitzler

Hurni

2022

IJGI

Self Cite

View full text Add to dashboard Cite

Since landforms composing land surface vary in their properties and appearance, their shaded reliefs also present different visual impression of the terrain. In this work, we adapt a U-Net so that it can recognize a selection of landforms and can segment terrain. We test the efficiency of 10 separate models and apply an ensemble approach, where all the models are combined to potentially outperform single models. Our algorithm works particularly well for block mountains, Prealps, valleys, and hills, delivering average precision and f1 values above 60%. Segmenting plateaus and folded mountains is more challenging, and their precision values are rather scattered due to smaller areas available for training. Mountains formed by erosion processes are the least recognized landform of all because of their similarities with other landforms. The highest accuracy of one of the 10 models is 65%, while the accuracy of the ensemble is 61%. We apply relief shading techniques that were found to be efficient regarding specific landforms within corresponding segmented areas and blend them together. Finally, we test the trained model with the best accuracy on other mountainous areas around the world, and it proves to work in other regions beyond the training area.

show abstract

“…These classification networks can also be used for the classification of types of maps (Zhou et al, 2018), or can be used for the selective omission in a road network (Zhou and Li, 2016). Segmentation networks can localize pixels that belong to a generalised object given the image of the map before generalisation (Courtial et al, 2020a;Feng et al, 2019;Jenny et al, 2020;Du et al, 2021). Generative adversarial networks (GANs) are another deep learning architecture that can be interesting for map generalisation, as they have shown potential for style transfer on maps (Kang et al, 2019), and were employed for building shape generalisation (Kang et al, 2020), although the results are not convincing yet.…”

Section: Deep Learning and Cartographymentioning

confidence: 99%

Generative Adversarial Networks to Generalise Urban Areas in Topographic Maps

Courtial

Touya

Zhang

2021

Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci.

View full text Add to dashboard Cite

Abstract. This article presents how a generative adversarial network (GAN) can be employed to produce a generalised map that combines several cartographic themes in the dense context of urban areas. We use as input detailed buildings, roads, and rivers from topographic datasets produced by the French national mapping agency (IGN), and we expect as output of the GAN a legible map of these elements at a target scale of 1:50,000. This level of detail requires to reduce the amount of information while preserving patterns; covering dense inner cities block by a unique polygon is also necessary because these blocks cannot be represented with enlarged individual buildings. The target map has a style similar to the topographic map produced by IGN. This experiment succeeded in producing image tiles that look like legible maps. It also highlights the impact of data and representation choices on the quality of predicted images, and the challenge of learning geographic relationships.

show abstract

Cartographic Relief Shading with Neural Networks

Abstract: Fig. 1. Shaded relief of the Caucasus Mountains created with a neural network trained with a manual relief shading of Switzerland.

Cited by 28 publications

References 62 publications

A Novel Data Augmentation Method to Enhance the Training Dataset for Road Extraction From Swiss Historical Maps

A Novel Data Augmentation Method to Enhance the Training Dataset for Road Extraction From Swiss Historical Maps

Terrain Segmentation Using a U-Net for Improved Relief Shading

Generative Adversarial Networks to Generalise Urban Areas in Topographic Maps

Contact Info

Product

Resources

About