Remote Sensing Image Inpainting with Generative Adversarial Networks

Kuznetsov, Andrey; Гашников, М. В.

doi:10.1109/isdfs49300.2020.9116347

Cited by 8 publications

(6 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…While the HLS version 2 product has been recently released and is expected to provide optical reflectance images every 2–3 days at 30 m resolution over the globe, the production process is still currently ongoing. These missing gaps can be potentially addressed through data fusion (e.g., Kuznetsov & Gashnikov, 2020; Li et al., 2022; Luo et al., 2018). Reanalysis and land data assimilation system (LDAS) data sets provide spatiotemporally continuous LST under all‐weather conditions (e.g., Quan et al., 2023; Rodell et al., 2004).…”

Section: Discussionmentioning

confidence: 99%

Improved 30‐m Evapotranspiration Estimates Over 145 Eddy Covariance Sites in the Contiguous United States: The Role of ECOSTRESS, Harmonized Landsat Sentinel‐2 Imagery, Climate Reanalysis, and Deep Neural Network Postprocessing

Rashid,

Tian

2024

Water Resources Research

View full text Add to dashboard Cite

This study developed and evaluated 30‐m daily evapotranspiration (ET) estimates using the Priestley‐Taylor Jet Propulsion Laboratory (PT‐JPL) model with ECOSTRESS, Moderate MODIS, harmonized Landsat Sentinel‐2 (HLS) imagery, ERA5‐Land reanalysis, and eddy covariance measurements. The new daily 30‐m ET showed significantly improved performance (overall, r = 0.8, RMSE = 1.736, KGE = 0.466) at 145 EC sites over contiguous United States compared to the current 70‐m ECOSTRESS ET (overall, r = 0.485, RMSE = 4.696, KGE = −0.841). A deep neural network postprocessing model trained with ET measurements from EC sites further improved the performance on test sites that were not used for model training (overall, r = 0.842, RMSE = 0.88, KGE = 0.792). The 30‐m ET estimation biases were significantly related to the biases in the upwelling longwave (RUL) and downwelling shortwave radiation (RDS) inputs, with ET estimates driven by MODIS radiation showing higher biases compared to those driven by ERA5‐Land radiation. The error diagnosis using random forest indicates that ET biases tend to be larger under higher ET estimates, and RUL and RDS were the primary contributors to the high bias at the higher ET ranges, with partial dependence plots revealing that the estimation biases tend to be higher under more humid environment, denser vegetation covers, and high net radiation conditions. In conclusion, higher spatial resolution satellite imagery of vegetation characteristics and higher temporal resolution radiation data, combined with continent‐wide EC measurements and deep learning, provided substantial added value for improving ET estimations at the field scale (30‐m).

show abstract

Section: Discussionmentioning

confidence: 99%

Improved 30‐m Evapotranspiration Estimates Over 145 Eddy Covariance Sites in the Contiguous United States: The Role of ECOSTRESS, Harmonized Landsat Sentinel‐2 Imagery, Climate Reanalysis, and Deep Neural Network Postprocessing

Rashid,

Tian

2024

Water Resources Research

View full text Add to dashboard Cite

show abstract

“…4 Kuznetsov et al obtained forged images from Google Earth and various aircraft interior paintings using GAN. 5 Working with a dataset containing a small number of images, Yarlagadda et al used support vector machines (SVMs) to classify the fake images that had been created using GAN. When the classification results are examined, it can be seen that the accuracy results range from 79.7% to 97.2%.…”

Section: Literature Reviewmentioning

confidence: 99%

“…Kuznetsov et al. obtained forged images from Google Earth and various aircraft interior paintings using GAN 5 . Working with a dataset containing a small number of images, Yarlagadda et al.…”

Section: Literature Reviewmentioning

confidence: 99%

Transferemble: a classification method for the detection of fake satellite images created with deep convolutional generative adversarial network

Sürücü

Di̇ri̇

2023

J. Electron. Imag.

View full text Add to dashboard Cite

As the number of government and commercial satellites increases, there is a large increase in Earth observation (EO) imagery. Using different locations and tools, images can be taken from more than one satellite. Manipulations are carried out on these images using a variety of different methods. The number of studies that have been done on the manipulation of EO images is very small. In recent years, generative adversarial networks (GANs), a major breakthrough in deep learning, have made it very easy to obtain fake images. In this study, scene-by-scene fake images were obtained with the deep convolutional GAN on the EuroSAT dataset, which is one of the EO image sets, and fake scene images were obtained from the original scenes. In this study, a dataset called RF-EuroSAT was created. It consists of 14 classes and 36,000 images. Five transfer learning models (VGG-16, DenseNet201, MobileNetV2, RegNetY320, and ResNet152V2) were used to classify this dataset. Using these models as feature extraction and ensemble models (XGBoost, CatBoost, and LightGBM) as classifiers, the classification process was performed using our proprietary transferemble model. The best result was obtained with an accuracy of 91.55% using our transferemble model, which is developed in a modular structure.

show abstract

“…Hence, deep learning has been introduced earlier to tackle this challenge such as in [2], which solved three missing information tasks in remote sensing data using a deep convolutional * Equal contribution author network combined with spatio-temporal information. Similarly, [3] used a generative approach for image inpainting. These images looked quite similar to the original version.…”

Section: Introductionmentioning

confidence: 99%

RSINet: Inpainting Remotely Sensed Images Using Triple GAN Framework

Kumar¹,

Tamboli²,

Pande³

et al. 2022

Preprint

View full text Add to dashboard Cite

We tackle the problem of image inpainting in the remote sensing domain. Remote sensing images possess high resolution and geographical variations, that render the conventional inpainting methods less effective. This further entails the requirement of models with high complexity to sufficiently capture the spectral, spatial and textural nuances within an image, emerging from its high spatial variability. To this end, we propose a novel inpainting method that individually focuses on each aspect of an image such as edges, colour and texture using a task specific GAN. Moreover, each individual GAN also incorporates the attention mechanism that explicitly extracts the spectral and spatial features. To ensure consistent gradient flow, the model uses residual learning paradigm, thus simultaneously working with high and low level features. We evaluate our model, alongwith previous state of the art models, on the two well known remote sensing datasets, Open Cities AI and Earth on Canvas, and achieve competitive performance. The code can be referred here: https://github.com/advaitkumar3107/RSINet.

show abstract

Remote Sensing Image Inpainting with Generative Adversarial Networks

Cited by 8 publications

References 16 publications

Improved 30‐m Evapotranspiration Estimates Over 145 Eddy Covariance Sites in the Contiguous United States: The Role of ECOSTRESS, Harmonized Landsat Sentinel‐2 Imagery, Climate Reanalysis, and Deep Neural Network Postprocessing

Improved 30‐m Evapotranspiration Estimates Over 145 Eddy Covariance Sites in the Contiguous United States: The Role of ECOSTRESS, Harmonized Landsat Sentinel‐2 Imagery, Climate Reanalysis, and Deep Neural Network Postprocessing

Transferemble: a classification method for the detection of fake satellite images created with deep convolutional generative adversarial network

RSINet: Inpainting Remotely Sensed Images Using Triple GAN Framework

Contact Info

Product

Resources

About