Multisenge: A Multimodal and Multitemporal Benchmark Dataset for Land Use/Land Cover Remote Sensing Applications

Wenger, Romain; Puissant, A.; Weber, Jonathan; Idoumghar, Lhassane; Forestier, Germain

doi:10.5194/isprs-annals-v-3-2022-635-2022

Cited by 10 publications

(11 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As detailed in Table III, MultiSenGE is composed of 5 urban classes and 9 natural classes. Additionally, to build this dataset exclusively images with a cloud cover less than 10% have been selected [48]. Similarly to PASTIS, no cloud masks are provided.…”

Section: A Data-setsmentioning

confidence: 99%

“…• The construction of a novel spatio-temporal architecture for SITS, named U-BARN To evaluate the performance of the proposed U-BARN architecture and the self-supervised training strategy, we conduct several experiments using the semantic segmentation downstream tasks defined by the labeled PASTIS data-set [19] and MultiSengGE [20]. First, the pre-trained U-BARN segmentation performances are compared with two end-toend trained architectures (U-TAE and U-BARN).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Self-Supervised Spatio-Temporal Representation Learning of Satellite Image Time Series

Dumeur,

Valero,

Inglada

2024

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

In this paper, a new self-supervised strategy for learning meaningful representations of complex optical Satellite Image Time Series (SITS) is presented. The methodology proposed, named U-BARN, a Unet-BERT spAtio-temporal Representation eNcoder, exploits irregularly sampled SITS. The designed architecture allows learning rich and discriminative features from unlabeled data, enhancing the synergy between the spatio-spectral and the temporal dimensions. To train on unlabeled data, a time series reconstruction pretext task inspired by the BERT strategy but adapted to SITS is proposed. A Sentinel-2 large-scale unlabeled data-set is used to pre-train U-BARN. During the pre-training, U-BARN processes annual time series composed of a maximum of 100 dates. To demonstrate its feature learning capability, representations of SITS encoded by U-BARN are then fed into a shallow classifier to generate semantic segmentation maps. Experimental results are conducted on a labeled crop data-set (PASTIS) as well as a dense land cover data-set (MultiSenGE). Two ways of exploiting U-BARN pretraining are considered: either U-BARN weights are frozen or fine-tuned. The obtained results demonstrate that representations of SITS given by the frozen U-BARN are more efficient for land cover and crop classification than those of a supervised-trained linear layer. Then, we observe that fine-tuning boosts U-BARN performances on MultiSenGE dataset. Additionally, we observe on PASTIS, in scenarios with scarce reference data, that the fine-tuning brings a significative performance gain compared to fully-supervised approaches. We also investigate the influence of the percentage of elements masked during pre-training on the quality of the SITS representation. Eventually, semantic segmentation performances show that the fully supervised U-BARN architecture reaches better performances than the spatiotemporal baseline (U-TAE) on both downstream tasks: crop and dense land cover segmentation.

show abstract

Section: A Data-setsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Self-Supervised Spatio-Temporal Representation Learning of Satellite Image Time Series

Dumeur,

Valero,

Inglada

2024

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

show abstract

“…As it focuses on seasonal changes in land cover, it may not be suitable for non-seasonal or long-term studies. 2 MultiSenGE [3] It aims to support the development of models that can generalize across different satellite sensors and is particularly useful for cross-sensor studies. Combining data from multiple sensors can be challenging and requires additional preprocessing.…”

Section: Data Descriptionmentioning

confidence: 99%

Sen-2 LULC: Land use land cover dataset for deep learning approaches

Sawant,

Garg,

Meshram

et al. 2023

Data in Brief

View full text Add to dashboard Cite

“…In this work, we have chosen to use MultiSenGE [16] which is a land use/land cover (LULC) dataset developed over the entire Grand-Est region in France. It contains 8,157 multitemporal and multimodal patches (256 × 256) cut from the Sentinel-1 and Sentinel-2 time series with represents 14 Sentinel-2 tiles.…”

Section: A Datasetsmentioning

confidence: 99%

Exploring inference of a land use and land cover model trained on MultiSenGE dataset

Wenger

Puissant

Weber

et al. 2023

2023 Joint Urban Remote Sensing Event (JURSE)

View full text Add to dashboard Cite

Land use and land cover and Urban Fabric (UF) mapping are very useful for urban modeling and simulation (growth, pollution, noise, micro-climate, mobility) in a context of global change. In recent years, due to the increase of Earth Observation data researchers built and shared datasets to the machine learning scientific community to apply and test their models for semantic segmentation. Few works have trained their deep learning model based on a geographic region and applied it to another geographical area of the same country. In this study, we explore the inference of a deep learning model pretrained on a multitemporal and multimodal dataset named 'MultiSenGE' dataset (based on a region in the East of France representing one fifth of the area of France) on five different cities over France (Toulouse, Dijon, Orléans, Lille, Rennes) away from the trained area. Results are encouraging and achieve F1-Score between 0.70 and 0.80 for the five urban fabric classes.

show abstract

Multisenge: A Multimodal and Multitemporal Benchmark Dataset for Land Use/Land Cover Remote Sensing Applications

Cited by 10 publications

References 21 publications

Self-Supervised Spatio-Temporal Representation Learning of Satellite Image Time Series

Self-Supervised Spatio-Temporal Representation Learning of Satellite Image Time Series

Sen-2 LULC: Land use land cover dataset for deep learning approaches

Exploring inference of a land use and land cover model trained on MultiSenGE dataset

Contact Info

Product

Resources

About