Toward a Collective Agenda on AI for Earth Science Data Analysis

Tuia, Devis; Roscher, Ribana; Wegner, Jan Dirk; Jacobs, Nathan; Zhu, Xiao Xiang; Camps‐Valls, Gustau

doi:10.1109/mgrs.2020.3043504

Cited by 45 publications

(17 citation statements)

References 79 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In Bejiga et al [27], by using a multi-date, multi-site and multi-spectral dataset, it is observed that the use of domain adaptation for a problem with more than one source of domain shift (i.e., images of different spatial locations and acquired at different time periods) increases the level of challenge in unsupervised domain adaptation approaches, affecting their performance. For remote sensing applications, this implies that there is a need to use domain-specific knowledge [41] that would help to reduce some degree of domain shift, hence augmenting the unsupervised domain adaptation approaches. Most applications in computer vision assume the scene layout between images is comparable with the differences stemming from the spectral characteristics, for example, an image of the street taken in the morning and in the afternoon.…”

Section: Discussionmentioning

confidence: 99%

Domain Adaptation for Semantic Segmentation of Historical Panchromatic Orthomosaics in Central Africa

Mboga

D’Aronco

Grippa

et al. 2021

IJGI

Self Cite

View full text Add to dashboard Cite

Multitemporal environmental and urban studies are essential to guide policy making to ultimately improve human wellbeing in the Global South. Land-cover products derived from historical aerial orthomosaics acquired decades ago can provide important evidence to inform long-term studies. To reduce the manual labelling effort by human experts and to scale to large, meaningful regions, we investigate in this study how domain adaptation techniques and deep learning can help to efficiently map land cover in Central Africa. We propose and evaluate a methodology that is based on unsupervised adaptation to reduce the cost of generating reference data for several cities and across different dates. We present the first application of domain adaptation based on fully convolutional networks for semantic segmentation of a dataset of historical panchromatic orthomosaics for land-cover generation for two focus cities Goma-Gisenyi and Bukavu. Our experimental evaluation shows that the domain adaptation methods can reach an overall accuracy between 60% and 70% for different regions. If we add a small amount of labelled data from the target domain, too, further performance gains can be achieved.

show abstract

Section: Discussionmentioning

confidence: 99%

Domain Adaptation for Semantic Segmentation of Historical Panchromatic Orthomosaics in Central Africa

Mboga

D’Aronco

Grippa

et al. 2021

IJGI

Self Cite

View full text Add to dashboard Cite

show abstract

“…During the demonstration of SyntEO, it could be shown how the approach can be utilised to establish fixed experiment environments and to generate highly adjustable dataset variants. Therewith, it was possible to make assumptions over the training process and model behaviour which is an important cornerstone towards explainable machine learning and interpretability of artificial intelligence, which is a current challenge in Earth observation and artificial intelligence [40]. The SyntEO approach starts a dialogue between machine and expert by using the ontology as a medium for exchange that both sides can interpret.…”

Section: Discussionmentioning

confidence: 99%

“…Lately, with the emergence of deep learning and especially its development in computer vision for image analysis [27], it became possible to extract complex spatio-temporal features from large amounts of remote sensing data. As a consequence, deep learning eventually established itself as an essential tool in Earth observation and gave new perspectives to geoscientific research [19,34,40,46].…”

Section: Introductionmentioning

confidence: 99%

SyntEO: Synthetic Data Set Generation for Earth Observation and Deep Learning -- Demonstrated for Offshore Wind Farm Detection

Hoeser¹,

Kuenzer²

2021

Preprint

View full text Add to dashboard Cite

With the emergence of deep learning in the last years, new opportunities arose in Earth observation research. Nevertheless, they also brought with them new challenges. The data-hungry training processes of deep learning models demand large, resource expensive, annotated datasets and partly replaced knowledge-driven approaches, so that model behaviour and the final prediction process became a black box. The proposed SyntEO approach enables Earth observation researchers to automatically generate large deep learning ready datasets and thus free up otherwise occupied resources. SyntEO does this by including expert knowledge in the data generation process in a highly structured manner. In this way, fully controllable experiment environments are set up, which support insights in the model training. Thus, SyntEO makes the learning process approachable and model behaviour interpretable, an important cornerstone for explainable machine learning. We demonstrate the SyntEO approach by predicting offshore wind farms in Sentinel-1 images on two of the worlds largest offshore wind energy production sites. The largest generated dataset has 90,000 training examples. A basic convolutional neural network for object detection, that is only trained on this synthetic data, confidently detects offshore wind farms by minimising false detections in challenging environments. In addition, four sequential datasets are generated, demonstrating how the SyntEO approach can precisely define the dataset structure and influence the training process. SyntEO is thus a hybrid approach that creates an interface between expert knowledge and data-driven image analysis.

show abstract

“…Deep learning in particular has led to a revolution in AI in recent years. Since taking off several years ago, deep learning in remote sensing has become a blooming research field ( Zhu et al, 2017 , Tuia et al, 2021 , Camps-Valls et al, 2021 ). Its huge potential in global urban mapping using EO data is ripe for discovery.…”

Section: Introductionmentioning

confidence: 99%

The urban morphology on our planet – Global perspectives from space

Zhu

Qiu

et al. 2022

Remote Sensing of Environment

Self Cite

View full text Add to dashboard Cite

Highlights Combined big and multisensory Earth observation data with deep learning. Revealed the first time detailed morphology of urban agglomerations across the globe. Statistical study of our results quantifies a global inequality in population density. Clustering of the global result identified seven unique urban morphological patterns. This global urban morphological dataset (So2Sat Global Urban LCZ) will be open access.

show abstract

Toward a Collective Agenda on AI for Earth Science Data Analysis

Cited by 45 publications

References 79 publications

Domain Adaptation for Semantic Segmentation of Historical Panchromatic Orthomosaics in Central Africa

Domain Adaptation for Semantic Segmentation of Historical Panchromatic Orthomosaics in Central Africa

SyntEO: Synthetic Data Set Generation for Earth Observation and Deep Learning -- Demonstrated for Offshore Wind Farm Detection

The urban morphology on our planet – Global perspectives from space

Contact Info

Product

Resources

About