Breaking Limits of Remote Sensing by Deep Learning From Simulated Data for Flood and Debris-Flow Mapping

Yokoya, Naoto; Yamanoi, Kazuki; He, Wei; Baier, Gerald; Adriano, Bruno; Miura, Hiroyuki; Oishi, Satoru

doi:10.1109/tgrs.2020.3035469

Cited by 30 publications

(12 citation statements)

References 66 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…With respect to the ground truth labels, classification is comparable for real and synthesized images, indicating that the quality of synthesized images is good enough not to confuse a network pre-trained on real data, which conversely should make it feasible to use synthesized images for training and real data during inference [27]. One notable exception are buildings in the GeoNRW dataset.…”

Section: Synthesizing Remote Sensing Imagerymentioning

confidence: 94%

“…In a similar vein, image synthesis could also help with the actual training of new neural networks by generating training data. As an example, numerical simulations of rain fall can produce flood and debris flow maps, which can then in turn be used to train machine learning models [27]. Such a method can be extended with our work by synthesizing images from said maps.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Synthesizing Optical and SAR Imagery From Land Cover Maps and Auxiliary Raster Data

Baier,

Deschemps,

Schmitt

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

We synthesize both optical RGB and SAR remote sensing images from land cover maps and auxiliary raster data using generative adversarial networks. In remote sensing many types of data, such as digital elevation models or precipitation maps, are often not reflected in land cover maps but still influence image content or structure. Including such data in the synthesis process increases the quality of the generated images and exerts more control on their characteristics.Our method fuses both inputs by spatially adaptive normalization layers, previously published as SPADE semantic image synthesis. In contrast to SPADE, these normalization layers are applied to a fullblown generator architecture consisting of encoder and decoder, to take full advantage of the information content in the auxiliary raster data.Our method successfully synthesizes medium (10 m) and high (1 m) resolution images, when trained with the corresponding dataset. We show the advantage of data fusion of land cover maps and auxiliary information using mean intersection over union, pixel accuracy and FID using pretrained U-Net segmentation models. Handpicked images exemplify how fusing information avoids ambiguities in the synthesized images. By slightly editing the input our method can be used to synthesize realistic changes, i.e., raising the water levels.The source code is available at https://github.com/gbaier/rs_img_synth and we published the newly created high-resolution dataset at https: //ieee-dataport.org/open-access/geonrw.

show abstract

Section: Synthesizing Remote Sensing Imagerymentioning

confidence: 94%

mentioning

confidence: 99%

Synthesizing Optical and SAR Imagery From Land Cover Maps and Auxiliary Raster Data

Baier,

Deschemps,

Schmitt

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Remote sensing (RS) images have been widely used in multiple applications related to earth observation, such as object detection and recognition [1]- [5], land-use or land-cover classification [6]- [12], disaster monitoring and management of natural resources [13], [14], among others. All these tasks require an accurate characterization of RS scenes, from which semantic concepts should be precisely captured.…”

Section: Introductionmentioning

confidence: 99%

Rotation-Invariant Deep Embedding for Remote Sensing Images

Kang

Fernández-Beltrán

Wang

et al. 2022

IEEE Trans. Geosci. Remote Sensing

View full text Add to dashboard Cite

Endowing convolutional neural networks (CNNs) with the rotation-invariant capability is important for characterizing the semantic contents of remote sensing (RS) images, since they do not have typical orientations. Most of the existing deep methods for learning rotation-invariant CNN models are based on the design of proper convolutional or pooling layers, which aim at predicting the correct category labels of the rotated RS images equivalently. However, few works have focused on learning rotationinvariant embeddings in the framework of deep metric learning for modeling the fine-grained semantic relationships among RS images in the embedding space. To fill this gap, we first propose a rule that the deep embeddings of rotated images should be closer to each other than those of any other images (including the images belonging to the same class). Then, we propose to maximize the joint probability of the leave-one-out image classification and rotational image identification. With the assumption of independence, such optimization leads to the minimization of a novel loss function composed of two This work was in part supported by the Ministry of Science, Innovation and Universities of Spain (RTI2018-098651-B-C54), the Valencian Government of Spain (GV/2020/167), FEDER-Junta de Extremadura (Ref. GR18060) and the European Union under the H2020 EOXPOSURE project (No. 734541).

show abstract

“…RS is deployed in many applications such as disaster mapping [19][20][21][22], environment monitoring [23,24], land Journal of Sensors use/cover mapping [25][26][27][28][29][30], and forest mapping [31,32]. Due to improvement of spatial and temporal resolution of satellite imagery and availability of synthetic aperture radar (SAR) dataset, disaster mapping based on RS data has been converted into a hot topic [33,34]. Meanwhile, algorithm development and computer sciences revolutionized the accuracy and time of different sensors' data processing in various fields such as RS, prediction of natural disaster, feature detection, and biomedical [35,36].…”

Section: Introductionmentioning

confidence: 99%

Fusion of the Multisource Datasets for Flood Extent Mapping Based on Ensemble Convolutional Neural Network (CNN) Model

Seydi

Saeidi²,

Kalantar

et al. 2022

Journal of Sensors

View full text Add to dashboard Cite

Floods, as one of the natural hazards, can affect the environment, damage the infrastructures, and threaten human lives. Due to climate change and anthropogenic activities, floods occur in high frequency all over the world. Therefore, mapping of the flood areas is of prime importance in disaster management. This research presents a novel framework for flood area mapping based on heterogeneous remote sensing (RS) datasets. The proposed framework fuses the synthetic aperture radar (SAR), optical, and altimetry datasets for mapping flood areas, and it is applied in three main steps: (1) preprocessing, (2) deep feature extraction based on multiscale residual kernel convolution and convolution neural network’s (CNN) parameter optimization by fusing the datasets, and (3) flood detection based on the trained model. This research exploits two large-scale area datasets for mapping the flooded areas in Golestan and Khuzestan provinces, Iran. The results show that the proposed methodology has a high performance in flood area detection. The visual and numerical analyses verify the effectiveness and ability of the proposed method to detect the flood areas with an overall accuracy (OA) higher than 98% in both study areas. Finally, the efficiency of the designed architecture was verified by hybrid-CNN and 3D-CNN methods.

show abstract

Breaking Limits of Remote Sensing by Deep Learning From Simulated Data for Flood and Debris-Flow Mapping

Cited by 30 publications

References 66 publications

Synthesizing Optical and SAR Imagery From Land Cover Maps and Auxiliary Raster Data

Synthesizing Optical and SAR Imagery From Land Cover Maps and Auxiliary Raster Data

Rotation-Invariant Deep Embedding for Remote Sensing Images

Fusion of the Multisource Datasets for Flood Extent Mapping Based on Ensemble Convolutional Neural Network (CNN) Model

Contact Info

Product

Resources

About