An attention mechanism based convolutional network for satellite precipitation downscaling over China

Jing, Yinghong; Lin, Liupeng; Li, Xinghua; Li, Tongwen; Shen, Huanfeng

doi:10.1016/j.jhydrol.2022.128388

Cited by 22 publications

(7 citation statements)

References 52 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Jing et al. [61] introduced an attention mechanism‐based CNN comprising a global cross‐attention and residual convolutional modules that consider the potential relationships between complicated characteristics.…”

Section: Related Workmentioning

confidence: 99%

Efficient face image super‐resolution with convenient alternating projection network

Chen

et al. 2023

IET Signal Processing

View full text Add to dashboard Cite

The existing deep learning‐based face super‐resolution techniques can achieve satisfactory performance. However, these methods often incur large computational costs, and deeper networks generate redundant features. Some lightweight reconstruction networks also present limited representation ability because they ignore the entire contour and fine texture of the face for the sake of efficiency. Here, the authors propose a convenient alternating projection network (CAPN) for efficient face super‐resolution. First, the authors design a novel alternating projection block cascaded convolutional neural network to alternately achieve content consistency and learn detailed facial feature differences between super‐resolution and ground‐truth face images. Second, the self‐correction mechanism enabled the convolutional layer to capture faithful features that facilitate adaptive reconstruction. Moreover, a convenient connection operation can reduce the generation of redundant facial features while maintaining accurate reconstruction information. Extensive experiments demonstrated that the proposed CAPN can effectively reduce the computational cost while achieving competitive qualitative and quantitative results compared to state‐of‐the‐art super‐resolution methods.

show abstract

Section: Related Workmentioning

confidence: 99%

Efficient face image super‐resolution with convenient alternating projection network

Chen

et al. 2023

IET Signal Processing

View full text Add to dashboard Cite

show abstract

“…Due to the complexity of spatiotemporal characteristics, downscaling remains a challenging and intricate problem. Over the past few decades, various downscaling techniques have been proposed, including simple downscaling, dynamical downscaling (Jing et al., 2022; F. Wang et al., 2021), and statistical downscaling (Fowler et al., 2007; Sharifi et al., 2019; Zhi et al., 2016). Among these, statistical downscaling exhibits a distinct advantage due to its high accuracy, excellent scalability, and lower computational resource requirements (Frei et al., 2003; Hagemann et al., 2004; L. Ji, Zhi, Schalge, et al., 2023; Kim & Barros, 2002; Mannig et al., 2013).…”

Section: Introductionmentioning

confidence: 99%

“…However, when U‐Net is employed for downscaling end‐to‐end tasks, the accuracy and practical effectiveness of the results can still be further improved through existing techniques. There are various attention mechanisms for image processing tasks (Hu et al., 2018; Mnih et al., 2014; Woo et al., 2018), and it has also found applications in downscaling (Gerges et al., 2022; Jing et al., 2022; Park et al., 2022). In theory, deeper networks have larger receptive fields, allowing them to integrate more information and potentially achieve better results.…”

Section: Introductionmentioning

confidence: 99%

Deep Learning for Daily 2‐m Temperature Downscaling

Ding,

Zhi,

Lyu

et al. 2024

Earth and Space Science

View full text Add to dashboard Cite

This study proposes a novel method, which is a U‐shaped convolutional neural network that combines non‐local attention mechanisms, Res2net residual modules, and terrain information (UNR‐Net). The original U‐Net method and the linear regression (LR) method are conducted as benchmarks. Generally, the UNR‐Net has demonstrated promise in performing a 10× downscaling for daily 2‐m temperature over North China with lead times of 1–7 days and shows superiority to the U‐Net and LR methods. To be specific, U‐Net and UNR‐Net demonstrate higher Nash‐Sutcliffe Efficiency coefficient values compared to LR by 0.052 and 0.077, respectively. The corresponding improvements in pattern correlation coefficient are 0.013 and 0.016, while the root mean square error values are higher by 0.22 and 0.338, respectively. Additionally, the structural similarity index metric is higher by 0.033 and lower by 0.015. Furthermore, regions with significant errors are primarily distributed in complex terrain areas such as the Taihang Mountains, where UNR‐Net exhibits noticeable improvements. In addition, the 12 components‐based error decomposition method is proposed to analyze the error source of different models. Generally, the smallest errors are observed during the summer season and the sequence error component is proven to be the main source error of 2‐m temperature forecasts. Furthermore, UNR‐Net consistently demonstrates the lowest errors among all 12 error components. Therefore, combining the numerical weather prediction model and deep learning method is very promising in downscaling temperature forecasts and can be applied to routine forecasting of other atmospheric variables in the future.

show abstract

“…Due to the complexity of spatiotemporal characteristics, downscaling remains a challenging and intricate problem. Over the past few decades, various downscaling techniques have been proposed, including simple downscaling, dynamical downscaling (Jing et al 2022;Wang et al 2021), and statistical downscaling (Sharifi et al, 2019;Fowler et al, 2007). Among these, statistical downscaling exhibits a distinct advantage due to its high accuracy, excellent scalability, and lower computational resource requirements (Kim & Barros 2002;Frei et al, 2003;Hagemann et al, 2004;Ji et al, 2023a;Mannig et al, 2013).…”

Section: Introductionmentioning

confidence: 99%

Deep Learning for Daily 2-m Temperature Downscaling

Ding,

Zhi,

Lyu

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

An attention mechanism based convolutional network for satellite precipitation downscaling over China

Cited by 22 publications

References 52 publications

Efficient face image super‐resolution with convenient alternating projection network

Efficient face image super‐resolution with convenient alternating projection network

Deep Learning for Daily 2‐m Temperature Downscaling

Deep Learning for Daily 2-m Temperature Downscaling

Contact Info

Product

Resources

About