Low-light-level image super-resolution reconstruction via deep learning network To solve these problems, we propose a multi-scale feature extraction (MSFE) network to realize super-resolution imaging in a low-light-level (LLL) ... B Wang,Y Zou,Y Li,... 被引量: 0发表: 2021年 Hyperspectral...
Wang S, Zhou T, Lu Y, Di H (2022) Detail-preserving transformer for light field image super-resolution. Proceedings of the AAAI Conference on Artificial Intelligence 36:2522–2530 Article Google Scholar Zamir SW, Arora A, Khan S, Hayat M, Khan FS, Yang M-H (2022) Restormer: Efficient...
Accurate Image Super-Resolution Using Very Deep Convolutional Networks笔记 摘要: (1)网络结构深:a significant improvement in accuracy 通过级联多层小的卷积核,上下文信息被高效利用 (2)深层网络存在的问题是收敛速度 采用学习残差和高的学习率 1.介绍 srcnn相对于传统学习方法,不需... ...
《AudioSR: Versatile Audio Super-resolution at Scale》(2023) GitHub: github.com/haoheliu/versatile_audio_super_resolution《Sight Beyond Text: Multi-Modal Training Enhances LLMs in Truthfulness and Ethics》(2023) GitHub: github.com/UCSC-VLAA/Sight-Beyond-Text [fig4]...
【AAAI2023】Ultra-High-Definition Low-Light Image Enhancement: A Benchmark and Transformer-Based Method代码:github.com/TaoWangzj/LL 这个论文首先构建了ultra-high definition low-light (UHD-LOL)数据集,然后提出了 Low-Light Transformer (LLFormer)。 LLFormer 的整体框架如下所示,可以看出和 Restormer 有...
Unsupervised Image Super-Resolution using Cycle-in-Cycle Generative Adversarial Networks论文阅读 1.存在的问题: 低分辨率图像作为输入时,由于噪声和模糊会进一步退化。这种复杂的设置使得有监督学习和准确的内核估计成为不可能。 2.解决方法: 采用无监督学习,使用没有配对数据,以生成式对抗网络(GAN)为基本组成... ...
His research interests include image restoration and image super-resolution. Jinjing Zhang received his PhD degree in 2019 and received her B.S. in 2014 from North University of China in Taiyuan, China. She is an assistant lecturer in North University of China. She is mainly majoring image ...
作者通过 attention 运算,计算一个3x3的相似性矩阵,给输入的特征进行加权。输入的三组特征里,强调重要的、抑制不重要的。Layer attention 的思路最早应该是在 【ECCV2020】Single Image Super-Resolution via a Holistic Attention Network 这个论文里出现(在任文琦老师报告里听到的)...
LLVIPLLVIP: A visible-infrared paired dataset for low-light visionlink RELLISURRELLISUR: A Real Low-Light Image Super-Resolution Datasetlink LSRWR2RNet: Low-light Image Enhancement via Real-low to Real-normal Network; 3170 paired images using the Nikon camera and 2480 paired images using the...
The commonly used loss functions in LLIE networks are also employed in image reconstruction networks for image super-resolution [73], image denoising [74], image detraining [75, 76, 77], and image deblurring [78]. Different from these versatile losses, the specially designed exposure loss for ...