SatGAN: Satellite Image Generation using Conditional Adversarial Networks

Shah, Mitt; Gupta, Manish; Thakkar, Priyank

doi:10.1109/iccict50803.2021.9510104

Cited by 5 publications

(2 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Furthermore, specific application cases of GANs have demonstrated their great potential in satellite and aerial image processing. One such well-known case involves automatically converting satellite images to maps and vice versa (Figure 3), which has been explored and presented in numerous research studies and projects such as [26], [27], [28], and [29], among many others.…”

Section: Related Workmentioning

confidence: 99%

Correction of Banding Errors in Satellite Images With Generative Adversarial Networks (GAN)

Paola,

Jesús,

Christian

et al. 2023

IEEE Access

View full text Add to dashboard Cite

This research proposes an innovative method for correcting banding errors in satellite images based on Generative Adversarial Networks (GAN). Small satellites are frequently launched into space to obtain images that can be used in scientific or military research, commercial activities, and urban planning, among other applications. However, its small cameras are more susceptible to radiometric, geometric errors, and other distortions caused by atmospheric interference. The proposed method was compared to the conventional correction technique using experimental data, showing the similar performance (92.64% and 90.05% accuracy, respectively). These experimental results suggest that generative models utilizing Artificial Intelligence (AI) techniques, specifically Deep Learning, are getting closer to achieving automatic correction close to conventional methods. Advantages of the GAN models include automating the task of correcting banding in satellite images, reducing the required time, and facilitating the processing without requiring prior technical knowledge in handling Geographic Information Systems (GIS). Potentially, this technique could represent a valuable tool for satellite image processing, improving the accuracy of the results and making the process more efficient. The research is particularly relevant to the field of remote sensing and can have practical applications in various industries.

show abstract

Section: Related Workmentioning

confidence: 99%

Correction of Banding Errors in Satellite Images With Generative Adversarial Networks (GAN)

Paola,

Jesús,

Christian

et al. 2023

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Prior works have also focused on the specific domain of remote sensing and satellite imagery. For example SatGAN introduce a pixel-level reconstruction loss to generate colorful and blur-free versions of the ground truth satellite images [20], other work from the United Nations Satellite Centre employs a conditional progressive GAN to generate missing tiles in an image with applications to flood detection [1], and ColorMapGAN propose training satellite images to match test images' spectral distributions to deal with the distribution shift [21].…”

Section: Introductionmentioning

confidence: 99%

Mask Conditional Synthetic Satellite Imagery

Le¹,

Reddy²,

Chen³

et al. 2023

Preprint

View full text Add to dashboard Cite

In this paper we propose a mask-conditional synthetic image generation model for creating synthetic satellite imagery datasets. Given a dataset of real high-resolution images and accompanying land cover masks, we show that it is possible to train an upstream conditional synthetic imagery generator, use that generator to create synthetic imagery with the land cover masks, then train a downstream model on the synthetic imagery and land cover masks that achieves similar test performance to a model that was trained with the real imagery. Further, we find that incorporating a mixture of real and synthetic imagery acts as a data augmentation method, producing better models than using only real imagery (0.5834 vs. 0.5235 mIoU). Finally, we find that encouraging diversity of outputs in the upstream model is a necessary component for improved downstream task performance. We have released code for reproducing our work on GitHub.

show abstract

Using Weight Reliability Masks on Imbalanced Datasets for Satellite Image Segmentation

Okhrimenko,

Kussul

2024

Cybern Syst Anal

View full text Add to dashboard Cite

SatGAN: Satellite Image Generation using Conditional Adversarial Networks

Cited by 5 publications

References 16 publications

Correction of Banding Errors in Satellite Images With Generative Adversarial Networks (GAN)

Correction of Banding Errors in Satellite Images With Generative Adversarial Networks (GAN)

Mask Conditional Synthetic Satellite Imagery

Using Weight Reliability Masks on Imbalanced Datasets for Satellite Image Segmentation

Contact Info

Product

Resources

About