检索结果-内蒙古大学图书馆

Weakly supervised nuclei segmentation based on pseudo label correction and

ARTIFICIAL INTELLIGENCE IN MEDICINE 2025年 164卷 103113-103113页

作者： Pan, Xipeng Song, Shilong Liu, Zhenbing Wang, Huadeng Li, Lingqiao Lu, Haoxiang Lan, Rushi Luo, Xiaonan Guilin Univ Elect Technol Guangxi Key Lab Image & G Intelligent Proc Guilin 541004 Guangxi Peoples R China Southern Med Univ Guangdong Prov Peoples Hosp Guangdong Acad Med Sci Dept Radiol Guangzhou 510080 Guangdong Peoples R China Guangdong Prov Peoples Hosp Guangdong Prov Key Lab Artificial Intelligence Med Guangzhou 510080 Guangdong Peoples R China Guilin Univ Elect Technol Int Joint Res Lab Spatiotemporal Informat & Intell Guilin 541004 Guangxi Peoples R China

Nuclei segmentation plays a vital role in computer-aided histopathology image analysis. Numerous fully supervised learning approaches exhibit amazing performance relying on pathological image with precisely annotations. Whereas, it is difficult and time-consuming in accurate manual labeling on pathological images. Hence, this paper presents a two-stage weakly supervised model including coarse and fine phases, which can achieve nuclei segmentation on whole slide images using only point annotations. In the coarse segmentation step, Voronoi diagram and K-means cluster results are generated based on the point annotations to supervise the training network. In order to cope with the different imaging conditions, an image adaptive clustering pseudo label algorithm is proposed to adapt the color distribution of different images. A Multi-scale Feature Fusion (MFF) module is designed in the decoder to better fusion the feature outputs. Additionally, to reduce the interference of erroneous cluster label, an Exponential Moving Average for cluster label Correction (EMAC) strategy is proposed. After the first step, an uncertainty estimation pseudo label denoising strategy is introduced to denoise Voronoi diagram and adaptive cluster label. In the fine segmentation step, the optimized labels are used for training to obtain the final predicted probability map. Extensive experiments are performed on MoNuSeg and TNBC public benchmarks, which demonstrate our proposed method is superior to other existing nuclei segmentation methods based on point labels. Codes are available at: https://***/SSL-droid/WNSPLCUD.

关键词： Nuclei segmentation Weak supervision Point annotation Pseudo label denoising

来源：评论

学校读者我要写书评

暂无评论

STDC-MA network for semantic segmentation

引用

IET image procESSING 2022年第14期16卷 3758-3767页

作者： Lei, Xiaochun Lu, Linjun Jiang, Zetao Gong, Zhaoting Lu, Chang Liang, Jiaming Xie, Junlin Guilin Univ Elect Technol GUET Sch Comp Sci & Informat Secur Guilin 541004 Guangxi Peoples R China Guilin Univ Elect Technol GUET Guangxi Key Lab Image & Graph Intelligent Proc Guilin 541004 Guangxi Peoples R China Guilin Univ Elect Technol GUET Sch Comp Sci & Informat Secur Guilin Guangxi Peoples R China

Semantic segmentation is applied extensively in autonomous driving and intelligent transportation with methods that highly demand spatial and semantic information. Here, an STDC-MA network is proposed to meet these demands. First, the STDC-Seg structure is employed in STDC-MA to ensure a lightweight and efficient structure. Subsequently, the feature alignment module is applied to understand the offset between high-level and low-level features, solving the problem of pixel offset related to upsampling on the high-level feature map. The approach implements the effective fusion between high-level features and low-level features. A hierarchical multiscale attention mechanism is adopted to reveal the relationship among attention regions from two different input sizes of one image. Through this relationship, regions receiving much attention are integrated into the segmentation results, thereby reducing the unfocused regions of the input image and improving the effective utilisation of multiscale information. STDC-MA maintains the segmentation speed as the STDC-Seg network while improving the segmentation accuracy of small objects. STDC-MA was verified on the validation dataset of Cityscapes. The segmentation result of STDC-MA attained 78.32% mIOU with the input of 0.5x scale, 4.92% higher than STDC-Seg.

关键词： segmentation speed STDC-MA network highly demand segmentation accuracy intelligent transportation geophysical image processing semantic segmentation image representation high-level features lightweight structure low-level features semantic information feature extraction STDC-Seg network deep learning (artificial intelligence) image segmentation STDC-Seg structure efficient structure hierarchical multiscale attention mechanism high-level feature map autonomous driving image classification feature alignment module image resolution

来源：评论

学校读者我要写书评

暂无评论

Model detection for grey forecasting model with polynomial term

引用

COMMUNICATIONS IN STATISTICS-SIMULATION AND COMPUTATION 2023年

作者： Ouyang, Zhi-yuan Wang, Meng Zhang, Tao Guangxi Univ Sci & Technol Sch Sci Liuzhou Peoples R China Guangxi Univ Sci & Technol Key Lab Intelligent Informat Proc & Graph Proc Liuzhou Peoples R China Guangxi Univ Sci & Technol Key Lab Intelligent Informat Proc & Graph Proc Liuzhou 545006 Peoples R China

A grey forecasting model with polynomial term which includes the traditional grey model (GM (1)), the nonhomogeneous grey model (NGM(1,1,k)) and the integer order grey model with a time power term (GM(1,1,t alpha)) as special cases is considered. In grey forecasting model with polynomial term, we need to balance the costs and benefits of using higher order term. The novel method of model detection to identify which orders of polynomial term are significant is proposed. Simulation studies prove the proposed methods can identify the true model and have good prediction performances. An empirical example about annual health expenditure in China is used to illustrate the accuracy of proposed method.

关键词： Annual health expenditure Grey forecasting model with polynomial term Iteration Model detection

来源：评论

学校读者我要写书评

暂无评论

Single Traffic image Deraining via Similarity-Diversity Model

引用

IEEE TRANSACTIONS ON intelligent TRANSPORTATION SYSTEMS 2024年第1期25卷 90-103页

作者： Li, Youxing Lan, Rushi Huang, Huiwen Zhou, Huiyu Liu, Zhenbing Pang, Cheng Luo, Xiaonan Guilin Univ Elect Technol Guangxi Key Lab Intelligent Proc Comp Image & Grap Guilin 541004 Peoples R China Univ Leicester Sch Informat Leicester England Guilin Univ Elect Technol Natl Local Joint Engn Res Ctr Satellite Nav & Loca Guilin 541004 Peoples R China

Single traffic image deraining technology based on deep learning is a vital branch of image preprocessing, which is of great help to intelligent monitoring systems and driving navigation system. It is well understood that established deraining methods are derived based on one specific imaging model, neglecting the underlying correlations between different weather models and thereby limiting the applicability of these standard methods in real scenarios. To ameliorate this issue, in this work, we first explore the inherent relationship between a rain model and the haze one established up to date. We discover that these two models experience similar degradations in the low-frequency components (i.e., similarity) but diverse degradations in the high-frequency areas (i.e., diversity). Based on these observations, we develop a Similarity-Diversity model to describe these characteristics. Afterwards, we introduce a novel deep neural network to restore the rain-free background embedding the similarity-diversity model, namely deep similarity-diversity network (DSDNet). Extensive experiments have been conducted to evaluate our proposed method that outperforms the other state of the art deraining techniques. On the other hand, we deploy the proposed algorithm with Google Vision API for object recognition, which also obtains satisfactory results both qualitatively and quantitatively.

关键词： Deep learning single traffic image deraining imaging model similarity diversity

来源：评论

学校读者我要写书评

暂无评论

Distributed Predefined-Time Control for Hybrid AC/DC Microgrid

引用

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS 2023年第8期70卷 8324-8333页

作者： Zhang, Yu Wang, Yan-Wu Liu, Xiao-Kang Yang, Wu Liang, Shu-Ming Key Lab Image Proc & Intelligent Control Wuhan 430074 Peoples R China Huazhong Univ Sci & Technol Sch Artificial Intelligence & Automat Wuhan 430074 Peoples R China Guangxi Univ Sch Elect Engn Nanning 530004 Peoples R China

Due to the popularization of distributed generators (DGs) and diversity of loads, hybrid microgrid, mixing AC/DC subgrids, has gradually become a popular research topic. However, there is lack of research results to achieve the control objectives of precise AC/DC bus voltage/frequency restoration and global power sharing among DGs in hybrid microgrid within a predefined time. In this article, a distributed predefined-time controller (DPTC) is designed to achieve the above objectives with unknown load power by employing a classK1 function and defining a unified error. The convergence time of bus voltage and output power of converters can be adjusted by a predefined parameter. Moreover, a distributed predefined-time observer is proposed to rely less on direct load measurements and guarantee the controller implementation in emergencies when load information is unavailable. Through hardwarein-the-loop experiment tests, the effectiveness of the DPTC is verified in scenarios of load change and plug-and-play. Comparative experiment indicates that the proposed controller has the potential for convenient adjustment of convergence time and the advantage of small overshoot.

关键词： Microgrids Voltage control Reactive power Hybrid power systems Frequency control Convergence Renewable energy sources AC/DC bus voltage regulation hybrid microgrid power sharing predefined-time control

来源：评论

学校读者我要写书评

暂无评论

A small object detection algorithm based on improved Faster RCNN

A small object detection algorithm based on improved Faster ...

引用

International Symposium on Artificial Intelligence and Robotics

作者： Tang, Liling Li, Fang Lan, Rushi Luo, Xiaonan Guilin Univ Elect Technol Guangxi Key Lab Image & Graph Intelligent Proc Guilin 541004 Peoples R China

ISBN: (数字)9781510646131

ISBN: (纸本)9781510646131;9781510646124

Faster RCNN, as a classical detection algorithm, still faces a huge challenge in detecting small objects. Therefore, we introduce a multi-scale auxiliary feature fusion strategy to make sure that each layer of features contains rich semantic and spatial information. Firstly, we introduce shallow features extracted by a multi-scale auxiliary feature network into the backbone network, as a way to ensure that there is sufficient spatial information for detecting small objects even for the deepest feature. Secondly, we design a fusion module to fuse the auxiliary feature and backbone feature. Finally, to make the object proposal boxes positioning more precise in the ROI classification and regression network, replace RoIPool with RoIAlign. Our experiments are conducted on PASCAL VOC and KITTI autopilot datasets. Compared with the conventional methods, the improved Faster RCNN algorithm has 2.48% and 3.09% improved in mean average precision on PASCAL VOC and KITTI datasets, respectively.

关键词： Faster RCNN small object detection multi-scale feature feature fusion

来源：评论

学校读者我要写书评

暂无评论

E-TPE: Efficient Thumbnail-Preserving Encryption for Privacy Protection in Visual Sensor Networks

引用

ACM TRANSACTIONS ON SENSOR NETWORKS 2024年第4期20卷 1-26页

作者： Zhao, Ruoyu Zhang, Yushu Wen, Wenying Lan, Rushi Xiang, Yong Nanjing Univ Aeronaut & Astronaut Coll Comp Sci & Technol Jiangjun Rd Nanjing 211106 Jiangsu Peoples R China Chinese Acad Sci Inst Informat Engn State Key Lab Informat Secur Shucun Rd Beijing 100085 Peoples R China Jiangxi Univ Finance & Econ Sch Informat Management Yuping Ave Nanchang 330032 Jiangxi Peoples R China Guilin Univ Elect Technol Guangxi Key Lab Image & Graph Intelligent Proc Guilin 541004 Guangxi Peoples R China Deakin Univ Sch Informat Technol Burwood Hwy Burwood Vic 3125 Australia

When visual sensor networks (VSNs) enter daily life in society, they not only bring great convenience but also cause people to worry about privacy. Traditional image encryption uses the snowflake effect to protect the privacy, but this can compromise the usability of directly browsing images captured by nodes in VSNs. Recently, thumbnail-preserving encryption (TPE) has been proposed to balance image usability and privacy. However, the Markov chain’s limited connectivity and high time cost, which result in low efficiency, may preclude its application in VSNs. Motivated by this, we propose a novel TPE scheme to protect the image collected in VSNs’ privacy efficiently. We first deeply study the bijective relationship between the three-pixel group and its rank, and then propose a new number method for both, based on which we design the efficient rank mapping. Subsequently, an efficient thumbnail-preserving image encryption via three-pixel rank mapping (E-TPE) is designed, achieving NR security and enhancing the Markov chain’s connectivity. The experiments demonstrate the scheme’s efficiency; that is, it is currently the quickest ideal-TPE scheme with NR security, and a good balance is achieved between usability and privacy.

关键词： Privacy protection usability security visual sensor networks

来源：评论

学校读者我要写书评

暂无评论

Context-aware Network for Pulmonary Nodule detection in CT images

Context-aware Network for Pulmonary Nodule detection in CT I...

引用

International Symposium on Artificial Intelligence and Robotics

作者： Zhang, Jiawei Lan, Rushi Pang, Cheng Luo, Xiaonan Guilin Univ Elect Technol Guangxi Key Lab Image & Graph Intelligent Proc Guilin 541004 Peoples R China

ISBN: (数字)9781510646131

ISBN: (纸本)9781510646131;9781510646124

Suffered from low resolution images and straightforward features, existing R-CNN based methods for pulmonary nodule detection usually fail in detecting objects with small scales. In this paper, we propose a novel context-aware network which takes the pulmonary regions and their neighbors for joint learning. The contextual cues of these regions reinforce each other, which is beneficial for the detection of small regions. Moreover, a set of redesigned anchors are used to adapted pulmonary nodules with various sizes. In order to avoid dilution by redundant samples specifying large nodules, a data enhancement strategy is implemented in the training stage by identifying hard samples. We test the proposed network on a dataset with 2000 lung images and demonstrate it perfouns well in detection of lung nodules with various sizes. The proposed method has 7% improved to the original Faster R-CNN.

关键词： lung cancer Convolutional Neural Networks CT images Faster R-CNN pulmonary nodule detection anchor

来源：评论

学校读者我要写书评

暂无评论

GIAE-Net: A gradient-intensity oriented model for multimodal lung tumor image fusion

引用

ENGINEERING SCIENCE AND TECHNOLOGY-AN INTERNATIONAL JOURNAL-JESTECH 2024年 54卷

作者： Zhou, Tao Liu, Long Lu, Huiling Niu, Yuxia Guo, Yujie Pan, Yunfeng Bao, Wenxing North Minzu Univ Sch Comp Sci & Engn Yinchuan 750021 Peoples R China Ningxia Med Univ Sch Med Informat & Engn Yinchuan 750004 Peoples R China North Minzu Univ Key Lab Image & Graph Intelligent Proc State Ethn Yinchuan 750021 Peoples R China

Multimodal medical image fusion plays an important role in medical clinical applications. However, gradient features and intensity features are not extracted inadequately in fusion methods. To solve the above problems, this paper proposes a Gradient-Intensity oriented Automatic Encode-Decode multimodal lung tumor image fusion model (GIAE-Net), there are two parallel branches in this network, one is gradient branch, and another is intensity branch. The main idea of this proposed network is as follows:Firstly, the gradient attention module (GAM) is designed to enhance the description ability of fine-grained spatial features by using gradient operators, so that the network can retain more edge details. Secondly, the intensity attention module (IAM) is constructed to enable the model to learn image intensity features, which can highlight the lesion region information. Thirdly, the gradient intensity fusion module (GIFM) and the feature flow fusion strategy are designed. It converts the fusion problem into the weight assignment problem of gradient and intensity features, and the image feature extraction and fusion are realized gradually. Finally, a new multimodal medical dataset of lung tumor PET-CT is established, which contains 2575 pairs of PET-CT images (PCLset). The experimental results on PCLset show that compared with other nine fusion models, the proposed model can achieve better fusion performance. In CT lung window images and PET images comparison experiment, SD, IE, AG, QAB/F, VIF and EI indexes are improved by 20.38 %,7.70 %,16.44 %, 21.90 %,11.52 % and 33.95 %, respectively.

关键词： Lung Tumor Multimodal medical image image fusion Gradient feature Intensity feature

来源：评论

学校读者我要写书评

暂无评论

CSTAN: A Deepfake Detection Network with CST Attention for Superior Generalization

引用

SENSORS 2024年第22期24卷 7101页

作者： Yang, Rui You, Kang Pang, Cheng Luo, Xiaonan Lan, Rushi Guilin Univ Elect Technol Guangxi Key Lab Image & G Intelligent Proc Guilin 541004 Peoples R China Guilin Univ Elect Technol Sch Comp Sci & Informat Secur Guilin 541004 Peoples R China Guilin Univ Elect Technol Int Joint Res Lab Spatio Temporal Informat Intelli Guilin 541004 Peoples R China

With the advancement of deepfake forgery technology, highly realistic fake faces have posed serious security risks to sensor-based facial recognition systems. Recent deepfake detection models mainly use binary classification models based on deep learning. Despite achieving high detection accuracy on intra-datasets, these models lack generalization ability when applied to cross-datasets. We propose a deepfake detection model named Channel-Spatial-Triplet Attention Network (CSTAN), which focuses on the difference between real and fake features, thereby enhancing the generality of the detection model. To enhance the feature-learning ability of the model for image forgery regions, we have designed the Channel-Spatial-Triplet (CST) attention mechanism, which extracts subtle local information by capturing feature channels and the spatial correlation of three different scales. Additionally, we propose a novel feature extraction method, OD-ResNet-34, by embedding ODConv into the feature extraction network to enhance its dynamic adaptability to data features. Trained on the FF++ dataset and tested on the Celeb-DF-v1 and Celeb-DF-v2 datasets, the experimental results show that our model has stronger generalization ability in cross-datasets than similar models.

关键词： deepfake detection attention mechanism detection model feature extraction

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：