Generative Adversarial Networks for Image Super-Resolution: A Survey

Tian, Chunwei; Zhang, Xuanyu; Lin, Jerry Chun-Wen; Zuo, Wangmeng; Zhang, Yanning

doi:10.48550/arxiv.2204.13620

Cited by 16 publications

(14 citation statements)

References 164 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Later on, methods utilizing Generative Adversarial Networks (GAN) 8 for super-resolution generation gradually became mainstream. By introducing adversarial loss, GANs 15,21 can generate more realistic and detail-rich high-resolution images.…”

Section: Relatedmentioning

confidence: 99%

Enhancing Maritime Surveillance Video Clarity through SWIN-ESR: A Super-Resolution Approach for Efficient and Precise Ship Safety Monitoring

Guo,

Chen,

Zhang

2024

Preprint

View full text Add to dashboard Cite

With advancements in hardware and the rapid development of the internet, the clarity of surveillance videos has significantly improved, allowing people to quickly understand incidents in a specific location without being physically present. This is especially crucial in maritime navigation, where there is a lack of the solid ground feeling of safety. High-definition surveillance videos transmitted from the sea become a medium for land-based entities to ensure maritime safety. However, due to the absence of base stations at sea, these videos must be transmitted via satellite communication, which is limited by bandwidth and efficiency. Under normal weather conditions, the upload speed for a movie is generally below 200kb/s, and even lower during bad weather, meaning it could take a whole day to transmit a two-hour movie. To rapidly transmit surveillance videos without changing hardware, the data is compressed during transmission, such as by reducing the resolution. However, there is a high demand for the clarity of these images or videos, necessitating the use of image super-resolution techniques at the receiving end to enhance resolution and reconstruct the original high-definition images or videos as closely as possible.To minimize the loss of video information or even eliminate it, a method based on video super-resolution using the SWIN-ESR network is proposed to restore the original video, achieving the goal of transmission. This paper will compare SWIN-ESR with SRCNN, SRGAN, ESRGAN, and Real-ESRGAN, and evaluate the models through a series of performance metrics. The experimental results demonstrate that SWIN-ESR performs well in various numerical aspects.

show abstract

Section: Relatedmentioning

confidence: 99%

Enhancing Maritime Surveillance Video Clarity through SWIN-ESR: A Super-Resolution Approach for Efficient and Precise Ship Safety Monitoring

Guo,

Chen,

Zhang

2024

Preprint

View full text Add to dashboard Cite

show abstract

“…Recent deep generative models only need hundreds of latent variables to obtain various highly realistic designs. The generative adversarial network and variational autoencoder (VAE) [20] families are the two most popular deep generative frameworks nowadays; GAN has the advantage of generating more realistic results [21]. A classical GAN frame involves two subnetworks, termed generator and discriminator.…”

Section: Related Workmentioning

confidence: 99%

GEO: A Computational Design Framework for Automotive Exterior Facelift

Huang

Chen

Yan

et al. 2023

ACM Trans. Knowl. Discov. Data

View full text Add to dashboard Cite

Exterior facelift has become an effective method for automakers to boost the consumers’ interest in an existing car model before it is redesigned. To support the automotive facelift design process, this study develops a novel computational framework – Generator, Evaluator, Optimiser (GEO), which comprises 3 components: a StyleGAN2-based design generator that creates different facelift designs; a convolutional neural network (CNN)-based evaluator that assesses designs from the aesthetics perspective; and a recurrent neural network (RNN)-based decision optimiser that selects designs to maximise the predicted profit of the targeted car model over time. We validate the GEO framework in experiments with real-world datasets and describe some resulting managerial implications for automotive facelift. Our study makes both methodological and application contributions. First, the generator’s mapping network and projection methods are carefully tailored to facelift where only minor changes are performed without affecting the family signature of the automobile brands. Second, two evaluation metrics are proposed to assess the generated designs. Third, profit maximisation is taken into account in the design selection. From a high-level perspective, our study contributes to the recent use of machine learning and data mining in marketing and design studies. To the best of our knowledge, this is the first study that uses deep generative models for automotive regional design upgrading and that provides an end-to-end decision-support solution for automakers and designers.

show abstract

“…While some methods attempt to reduce time through feature space processing or reducing sampling steps, these often require additional operations with limited improvement. [5][6][7] Even with these optimizations, generating a small image still takes several seconds, making diffusion models over 100 times slower than Generative Adversarial Networks (GANs). To address this, Diffu-sionGAN 8 was proposed, merging Diffusion and GAN into one system, achieving breakthrough progress in inference speed.…”

Section: Introductionmentioning

confidence: 99%

Contour wavelet diffusion: A fast and high‐quality image generation model

Ding,

Zhu,

Zou

2024

Computational Intelligence

View full text Add to dashboard Cite

Diffusion models can generate high‐quality images and have attracted increasing attention. However, diffusion models adopt a progressive optimization process and often have long training and inference time, which limits their application in realistic scenarios. Recently, some latent space diffusion models have partially accelerated training speed by using parameters in the feature space, but additional network structures still require a large amount of unnecessary computation. Therefore, we propose the Contour Wavelet Diffusion method to accelerate the training and inference speed. First, we introduce the contour wavelet transform to extract anisotropic low‐frequency and high‐frequency components from the input image, and achieve acceleration by processing these down‐sampling components. Meanwhile, due to the good reconstructive properties of wavelet transforms, the quality of generated images can be maintained. Second, we propose a Batch‐normalized stochastic attention module that enables the model to effectively focus on important high‐frequency information, further improving the quality of image generation. Finally, we propose a balanced loss function to further improve the convergence speed of the model. Experimental results on several public datasets show that our method can significantly accelerate the training and inference speed of the diffusion model while ensuring the quality of generated images.

show abstract

Generative Adversarial Networks for Image Super-Resolution: A Survey

Cited by 16 publications

References 164 publications

Enhancing Maritime Surveillance Video Clarity through SWIN-ESR: A Super-Resolution Approach for Efficient and Precise Ship Safety Monitoring

Enhancing Maritime Surveillance Video Clarity through SWIN-ESR: A Super-Resolution Approach for Efficient and Precise Ship Safety Monitoring

GEO: A Computational Design Framework for Automotive Exterior Facelift

Contour wavelet diffusion: A fast and high‐quality image generation model

Contact Info

Product

Resources

About