2021 IEEE Winter Conference on Applications of Computer Vision (WACV) 2021
DOI: 10.1109/wacv48630.2021.00040
|View full text |Cite
|
Sign up to set email alerts
|

Hierarchical Generative Adversarial Networks for Single Image Super-Resolution

Abstract: Pathology image are essential for accurately interpreting lesion cells in cytopathology screening, but acquiring high-resolution digital slides requires specialized equipment and long scanning times. Though superresolution (SR) techniques can alleviate this problem, existing deep learning models recover pathology image in a black-box manner, which can lead to untruthful biological details and misdiagnosis. Additionally, current methods allocate the same computational resources to recover each pixel of patholog… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

0
5
0

Year Published

2022
2022
2024
2024

Publication Types

Select...
6
1
1

Relationship

0
8

Authors

Journals

citations
Cited by 15 publications
(5 citation statements)
references
References 64 publications
0
5
0
Order By: Relevance
“…The weight of the adversarial loss (λgan$\lambda _{gan}$) adjusts the degree of learning of the the distribution of the unpaired high‐dose CT images. If λgan$\lambda _{gan}$ is relatively higher than λobj$\lambda _{obj}$, the network may generate some unwanted artifacts because of the behavior of the GAN‐based methods 44 . Thus, we need to optimize the value of λobj$\lambda _{obj}$ to generate realistically denoised CT images.…”
Section: Resultsmentioning
confidence: 99%
See 1 more Smart Citation
“…The weight of the adversarial loss (λgan$\lambda _{gan}$) adjusts the degree of learning of the the distribution of the unpaired high‐dose CT images. If λgan$\lambda _{gan}$ is relatively higher than λobj$\lambda _{obj}$, the network may generate some unwanted artifacts because of the behavior of the GAN‐based methods 44 . Thus, we need to optimize the value of λobj$\lambda _{obj}$ to generate realistically denoised CT images.…”
Section: Resultsmentioning
confidence: 99%
“…If 𝜆 gan is relatively higher than 𝜆 obj , the network may generate some unwanted artifacts because of the behavior of the GAN-based methods. 44 Thus, we need to optimize the value of 𝜆 obj to generate realistically denoised CT images. Among two tested 𝜆 obj values (10.0 and 100.0), we observed that the denoised images with 𝜆 obj = 10.0 exhibit more edge-preserving results with the smallest LPIPS value.…”
Section: Optimized Value For Objective Weightmentioning
confidence: 99%
“…In addition, the automation facilitated by AI-OCR reduces the risk of human error. Automated extraction and input processes significantly mitigate the risk of mistakes that can result from manual data entry, leading to more reliable and accurate data [48].…”
Section: The Role Of Ai-ocr In Risk Reduction Spotlighting Real-time ...mentioning
confidence: 99%
“…Therefore, Chen et al proposed an HSRGAN network, which is divided into two modules. The first module divides the network into multiple branches, and the size of the convolution kernel of each branch is different [13]. At the end of the module, a feature fusion network is used to fuse these features.…”
Section: Image Super-resolutionmentioning
confidence: 99%