A5100654490- Image Enhancement Techniques
- Advanced Image Processing Techniques
- Advanced Vision and Imaging
- Cell Image Analysis Techniques
- Bone Metabolism and Diseases
- Computational Drug Discovery Methods
- Image Processing Techniques and Applications
- Cytokine Signaling Pathways and Interactions
- Peroxisome Proliferator-Activated Receptors
- Natural Language Processing Techniques
- Advanced Neural Network Applications
- Spine and Intervertebral Disc Pathology
- Bioinformatics and Genomic Networks
- Spinal Fractures and Fixation Techniques
- Acupuncture Treatment Research Studies
- vaccines and immunoinformatics approaches
- Estrogen and related hormone effects
- Free Radicals and Antioxidants
- Speech and dialogue systems
- Advanced Image and Video Retrieval Techniques
- Bone health and osteoporosis research
- Medical Image Segmentation Techniques
- Oral Health Pathology and Treatment
- Metal complexes synthesis and properties
- AI in cancer detection
Xi'an University of Technology
2022-2024
Shandong University of Traditional Chinese Medicine
2022-2024
University of Science and Technology of China
2022-2023
Shandong Normal University
2023
Baidu (China)
2022
Nanjing University
1999
Accurate vocal cord leukoplakia classification is critical for the individualized treatment and early detection of laryngeal cancer. Numerous deep learning techniques have been proposed, but it unclear how to select one apply in tasks. This article introduces reliably evaluates existing models classification.We created white light narrow band imaging (NBI) image datasets which were classified into six classes: normal tissues (NT), inflammatory keratosis (IK), mild dysplasia (MiD), moderate...
With economic expansion having moderated to a “new normal” pace, the eastern coastal provinces have been given new historical task of high-quality development and become window frontier China’s development. By designing optimizing an index system levels using entropy-TOPSIS method, study selected 21 indicators, include vitality, residents’ living standards, innovation efficiency green development, took as example characteristics spatial-temporal variations in level from 2010 2020. Then, used...
Identifying the interactions between T-cell receptor (TCRs) and human antigens is a crucial step in developing new vaccines, diagnostics, immunotherapy. Current methods primarily focus on learning binding patterns from known TCR repertoires by using sequence information alone without considering specificity of or exogenous peptides that have not appeared training set. Furthermore, spatial structure plays critical role immune studies immunotherapy, which should be addressed properly...
Modified Duhuo Jisheng Decoction (MDHJSD) is a traditional Chinese medicine prescription for the treatment of osteoporosis (OP), but its mechanism action has not yet been clarified. This study aims to explore MDHJSD in OP through combination network pharmacology analysis and experimental verification.The active ingredients corresponding targets were acquired from Traditional Medicine System Pharmacology (TCMSP) database. OP-related databases, including Genecards, OMIM, Drugbank, CTD, PGKB....
Recent advancements of artificial intelligence based on deep learning algorithms have made it possible to computationally predict compound-protein interaction (CPI) without conducting laboratory experiments. In this manuscript, we integrated a graph attention network (GAT) for compounds and long short-term memory neural (LSTM) proteins, used end-to-end representation both proposed algorithm, CPGL (CPI with GAT LSTM) optimize the feature extraction from proteins improve model robustness...
目的 现有大多数低照度图像增强算法会放大噪声,且用于极低照度图像时会出现亮度提升不足、色彩失真等问题。为此,提出一种基于 Retinex(retina cortex)的增强与去噪方法。方法 为了增强极低照度图像,首先利用暗通道先验原理估计场景的全局光照,若光照低于 0.5,对图像进行初始光照校正;其次,提出一种 Retinex 顺序分解模型,使低照度图像中的噪声均体现在反射分量中,基于分解结果,利用 Gamma 校正求取增强后的噪声图像;最后,提出一种基于内外双重互补先验约束的去噪机制,利用非局部自相似性原理为反射分量构建内部先验约束,基于深度学习,为增强后的噪声图像构建外部先验约束,使内外约束相互制约。结果 将本文算法与 6 种算法比较,在 140 幅普通低照度图像和 162 幅极低照度图像上(有正常曝光参考图像)进行主观视觉和客观指标评价比较,结果显示本文方法在亮度提升、色彩保真及去噪方面均有明显优势,对于普通低照度图像,BTMQI(blind tone-mappedquality index)和 NIQE(natural image quality...
目的 现有的低照度图像增强算法常存在局部区域欠增强、过增强及色彩偏差等情况,且对于极低照度图像增强,伴随着噪声放大及细节信息丢失等问题。对此,提出了一种基于照度与场景纹理注意力图的低光图像增强算法。方法 首先,为了降低色彩偏差对注意力图估计模块的影响,对低光照图像进行了色彩均衡处理;其次,试图利用低照度图像最小通道约束图对正常曝光图像的照度和纹理进行注意力图估计,为后续增强模块提供信息引导;然后,设计全局与局部相结合的增强模块,用获取的照度和场景纹理注意力估计图引导图像亮度提升和噪声抑制,并将得到的全局增强结果划分成图像块进行局部优化,提升增强性能,有效避免了局部欠增强和过增强的问题。结果 将本文算法与2种传统方法和4种深度学习算法比较,主观视觉和客观指标均表明本文增强结果在亮度、对比度以及噪声抑制等方面取得了优异的性能。在VV(Vasileios Vonikakis)数据集上,本文方法的BTMQI(blind tone-mapped quality index)和NIQMC(no-reference image metric for contrast...
This article uses microscopy images obtained from diverse anatomical regions of macaque brain for neuron semantic segmentation. The complex structure brain, the large intra-class staining intensity difference within class, small inter-class between and tissue unbalanced dataset increase difficulty To address this problem, we propose a multiscale segmentation- error-guided iterative convolutional neural network (MSEG-iCNN) to improve segmentation performance in major brain. After evaluating...
China is an argicultural country, and apples are one of the most prominentfruits with high production volume. Apple leaf diseases major factors that affect both quantity quality apple production. Usually,complex background small disease area in images make it difcult to accurately detect disease. In this paper, we propose improvedYOLOv7 method, called YOLOv7-PFA (YOLOv7 Parameter-Free Attention Mechanism) . Based on YOLOv7 algorithm, improve themodel’s detection ability for targets enhance...
To achieve more accurate prediction, advanced semantic segmentation methods are explored in the way of context modeling. Images real scenes usually contain multi-scale objects and contents. The feature propagation form convolution networks is very important to capture obtain segmentation. This paper proposes a novel pattern information flow aggregation rich features expression, called Cross Aggregation Module (CRA), In CRA, flows different scales transmitted module (CAM) through...
Representation of language is the first and critical task for Natural Language Understanding (NLU) in a dialogue system. Pretraining, embedding model, fine-tuning intent classification slot-filling are popular well-performing approaches but time consuming inefficient low-resource languages. Concretely, out-of-vocabulary transferring to different languages two tough challenges multilingual pretrained cross-lingual models. Furthermore, quality-proved parallel data necessary current frameworks....