检索结果-内蒙古大学图书馆

A Study on Optimization and Augmentation Techniques for Improving the Performance of CADx System for Cat Skin Disease

引用

Transactions of the Korean Institute of Electrical Engineers 2025年第1期74卷 135-141页

作者： Won, Hyeong-Sik Park, Jae-Beom Cho, Hyun-Chong Dept. of Electronics Engineering Interdisciplinary Graduate Program for BIT Medical Convergence Dept. of Data Science Kangwon National University Korea Republic of Dept. of Interdisciplinary Graduate Program for BIT Medical Convergence Kangwon National University Korea Republic of

In modern society, the number of households raising pets is increasing. As pet ownership increases, the cost of treating companion cats is also rising, with a significant portion of these costs going toward the treatment of skin diseases. Skin diseases are among the common ailments in pets, and swift action is required if they occur in cats. However, early lesions lack distinctive characteristics, making accurate diagnosis difficult. Therefore, this study proposes a CADx(Computer-Aided Diagnosis) system that classifies images of cat skin conditions into inflammatory lesions, non-inflammatory lesions, and normal images using a dataset of cat skin diseases. We selected the ConvNeXt model, based on CNN, which can learn regional information and features. To learn various patterns of skin lesions, we applied Mixup, an image augmentation technique. When Mixup was applied, the model accuracy was 0.8679, showing a 3.35% improvement compared to the original dataset. Additionally, Lookahead Optimizer was applied to ensure stable learning of Mixup. When both Mixup and Lookahead Optimizer were applied, the model accuracy was 0.8741, showing a high improvement of 3.97% compared to the original dataset. Therefore, the Lookahead Optimizer helps with the stable training of mixup, which can improve the performance of the model. Copyright © The Korean Institute of Electrical Engineers.

关键词： Dermatology

来源：评论

学校读者我要写书评

暂无评论

Development of a Deep Learning-Based System for Cat Skin Disease Classification and Grad-CAM Visualization

引用

Transactions of the Korean Institute of Electrical Engineers 2025年第2期74卷 339-344页

作者： Won, Hyeong-Sik Cho, Hyun-Chong Dept. of Interdisciplinary Graduate Program for BIT Medical Convergence Kangwon National University Korea Republic of Dept. of Electronics Engineering and Interdisciplinary Graduate Program for BIT Medical Convergence Dept. of Data Science Kangwon National University Korea Republic of

Skin diseases in companion cats can worsen if not treated promptly, and this can increase the financial burden on pet owners. To prevent this, early and accurate diagnosis is essential. This study introduces a deep learning-based Computer-Aided Diagnosis (CADx) system designed to classify cat skin diseases into non-inflammatory and inflammatory lesions by comparing them with normal images. The system employs the EfficientNetV2 model and incorporates image augmentation techniques like AutoAugment and AugMix to enhance classification performance. The study results indicate that the developed model achieved an 84.68% accuracy for non-inflammatory lesions, reflecting a 10.06% improvement, and a 97.19% accuracy for inflammatory lesions, reflecting a 2.45% improvement. Furthermore, we applied Grad-CAM to visualize the regions of interest in the images, offering veterinarians critical insights into the location and characteristics of the lesions. This system has the potential to significantly improve the precision of diagnosing skin diseases in companion cats, thereby supporting better veterinary care. Copyright © The Korean Institute of Electrical Engineers.

关键词： Veterinary medicine

来源：评论

学校读者我要写书评

暂无评论

Comparative Model Evaluation of Lightweight Transformer-Based and CNN-Based Architechture for Degenerative Lumbar Spine Classification

Comparative Model Evaluation of Lightweight Transformer-Base...

引用

2025 International Conference on Advancement in data science, E-learning and Information System, ICADEIS 2025

作者： Limicia, Jennifer Ardelia Chandra, Wilbert Gunawan, Alexander Agung Santoso Tedjasulaksana, Jeffrey Junior School of Computer Science Bina Nusantara University Data Science Program Computer Science Department Jakarta11480 Indonesia School of Computer Science Bina Nusantara University Computer Science Department Jakarta11480 Indonesia

ISBN: (纸本)9798331513320

Low back pain is a leading cause of disability globally, is often associated with degenerative lumbar spine conditions. Accurate diagnosis of these conditions is critical but challenging due to the subjective nature of MRI interpretation and the weak correlation between imaging findings and symptoms. Thus, the purpose of this study is to evaluate the performance of CNN-based architectures (VGG-16, EfficientNetB0, EfficientNetV2) and transformer-based architecture specifically Vision Transformers (ViT) in classifying lumbar spine conditions as normal, moderate, or severe. Using the RSNA 2024 Lumbar Spine Degenerative Classification dataset. The models are evaluated using the ROC-AUC and PRAUC as the performance metrics due to the dataset imbalance. Results indicate that EfficientNetB0 achieved the highest overall performance, with an average ROC-AUC of 0.784 and PR-AUC of 0.528, demonstrating strong adaptability to imbalanced datasets. EfficientNetV2 also performed competitively, while VGG-16 showed moderate effectiveness. The Vision Transformer (ViT), however, underperformed due to its reliance on larger datasets and challenges in capturing fine-grained spatial features. The findings highlight the potential of EfficientNet-based models for accurate and efficient lumbar spine diagnostics. This study underscores the potential of advanced deep learning approaches in improving diagnostic workflows for degenerative lumbar spine conditions. © 2025 IEEE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

An Analysis of OpenSeeD for Video Semantic Labeling

引用

Journal of Computer and Communications 2025年第1期13卷 59-71页

作者： Jenny Zhu NYU Center for Data Science GSTEM Program New York NY USA

Semantic segmentation is a core task in computer vision that allows AI models to interact and understand their surrounding environment. Similarly to how humans subconsciously segment scenes, this ability is crucial for scene understanding. However, a challenge many semantic learning models face is the lack of data. Existing video datasets are limited to short, low-resolution videos that are not representative of real-world examples. Thus, one of our key contributions is a customized semantic segmentation version of the Walking Tours dataset that features hour-long, high-resolution, real-world data from tours of different cities. Additionally, we evaluate the performance of open-vocabulary, semantic model OpenSeeD on our own custom dataset and discuss future implications.

关键词： Semantic Segmentation Detection Labeling OpenSeeD Open-Vocabulary Walking Tours dataset Videos

来源：评论

学校读者我要写书评

暂无评论

Adaptive Input Sampling: A Novel Approach for Efficient Object Detection in High Resolution Traffic Monitoring Images

引用

IEEE Access 2025年 13卷 73003-73021页

作者： Gebre, Tewodros Syum Wasehun, Eden Tsehaye Dorbu, Freda Elikem Agboola, Gazali Oluwasegun Hashemi-Beni, Leila Karimoddini, Ali NC AandT State University Applied Science and Technology Program GreensboroNC United States NC AandT State University Department of Computational Data Science and Engineering GreensboroNC United States NC AandT State University Geomatics Program Department of Built Environment GreensboroNC United States NC AandT State University Department of Electrical and Computer Engineering GreensboroNC United States

Real-time traffic monitoring is crucial for efficient transportation management, but it poses significant computational challenges, particularly when dealing with high-resolution videos. Most modern traffic monitoring systems employ object detection models followed by object tracking models, which require processing every frame of the video footage to detect and track vehicles. However, even efficient object detection models are computationally expensive, making it difficult to process high-resolution videos in real-time. To address this challenge, we introduce an Adaptive Input Sampler (AIS) as a preprocessing module that selectively activates the object detector only when vehicles are present in the scene. The AIS targets the area of the frame where vehicles are detected and samples that area without compressing the sample, thereby reducing the computational load on the object detector. By doing so, the AIS enables significant computational savings, especially when fewor no vehicles are present in the monitored scene. Our experiments demonstrate that the AIS achieves 96.40% traffic counting accuracy and 95.60% classification accuracy, while reducing the data size for analysis by up to 79.00%. AIS presents a promising solution to optimize the volume of input data in traffic monitoring applications, improve computational efficiency, and speed up inference results. © 2013 IEEE.

关键词： Footage counters

来源：评论

学校读者我要写书评

暂无评论

Advancing Precision Agriculture: Machine Learning-Enhanced GPR Analysis for Root-Zone Soil Moisture Assessment in Mega Farms

IEEE Transactions on AgriFood Electronics

引用

IEEE Transactions on AgriFood Electronics 2025年第1期3卷 98-109页

作者： Namdari, Himan Moradikia, Majid Zekavat, Seyed Askari, Radwin Mangoubi, Oren Petkie, Doug Data Science Program WorcesterMA01609 United States Michigan Technological University Department of Geological and Mining Engineering and Sciences HoughtonMI48105 United States Physics Department WorcesterMA01609 United States

In this article, we investigate an intelligent ground penetrating radar (GPR) that facilitates root-zone soil moisture estimation, a key parameter in precision agriculture. To create an intelligent GPR, we must train machine learning (ML) methods applied to the GPR-received signal. This process requires a large number of labeled GPR data that would be time-consuming and labor-intensive if created via field measurements. This article uses gprMAX software to emulate drone-coupled GPR received signal to generate large-scale data for training ML models. The data are created via a 1.5 GHz Ricker waveform considering a three-layer soil consistent with a realistic soil horizon model. The approach is structured as follows: first, we generate a synthetic dataset using gprMAX. Feature engineering techniques are then employed to extract meaningful components from the GPR signals, followed by a rigorous selection process to identify the most effective ML model for soil moisture prediction. Finally, we validate our model by integrating synthetic data with real GPR data collected at the SoilX lab at Worcester Polytechnic Institute, enhancing prediction accuracy and generalization capability. Our proposed model achieves an overall average root-mean-squared error of 0.5%, and 1.56 cm for moisture and depth estimations, respectively. The proposed intelligent GPR, when installed on a drone, enables high horizontal (e.g., 10 m) and vertical (e.g., 1.5 cm) resolution and high penetration depth (beyond 2 m) megafarm root-zone 3-D moisture map creation. Thus, it offers much higher capabilities when compared to traditional methods, such as synthetic aperture radar and satellite imaging. These results facilitate efficient farming practices, such as optimizing irrigation models, for better crop yields and food security. © 2023 IEEE.

关键词： Soil moisture

来源：评论

学校读者我要写书评

暂无评论

Hidden or Inferred: Fair Learning-to-Rank with Unknown Demographics 24

Hidden or Inferred: Fair Learning-to-Rank with Unknown Demog...

引用

Proceedings of the 2024 AAAI/ACM Conference on AI, Ethics, and Society

作者： Oluseun Olulana Kathleen Cachel Fabricio Murai Elke Rundensteiner Data Science Program Worcester Polytechnic Institute Worcester MA Data Science Program Worcester Polytechnic Institute Worcester MA and Computer Science Department Worcester Polytechnic Institute Worcester MA

As learning-to-rank models are increasingly deployed for decision-making in areas with profound life implications, the FairML community has been developing fair learning-to-rank (LTR) models. These models rely on the availability of sensitive demographic features such as race or sex. However, in practice, regulatory obstacles and privacy concerns protect this data from collection and use. As a result, practitioners may either need to promote fairness despite the absence of these features or turn to demographic inference tools to attempt to infer them. Given that these tools are fallible, this paper aims to further understand how errors in demographic inference impact the fairness performance of popular fair LTR strategies. In which cases would it be better to keep such demographic attributes hidden from models versus infer them? We examine a spectrum of fair LTR strategies ranging from fair LTR with and without demographic features hidden versus inferred to fairness-unaware LTR followed by fair re-ranking. We conduct a controlled empirical investigation modeling different levels of inference errors by systematically perturbing the inferred sensitive attribute. We also perform three case studies with real-world datasets and popular open-source inference methods. Our findings reveal that as inference noise grows, LTR-based methods that incorporate fairness considerations into the learning process may increase bias. In contrast, fair re-ranking strategies are more robust to inference errors. All source code, data, and experimental artifacts of our experimental study are available here: https://***/sewen007/***

关键词：

来源：评论

学校读者我要写书评

暂无评论

Group-Theoretical Perspective of Tree Editing

Group-Theoretical Perspective of Tree Editing

引用

IEEE International Conference on Consumer Electronics (ICCE)

作者： Tse-Yu Lin Data Science Degree Program National Taiwan University Taipei Taiwan

ISBN: (数字)9798331521165

ISBN: (纸本)9798331521172

Tree Editing Distance is a widely applied quantity for measuring the similarity between hierarchical data structures, particularly trees. This paper reinterprets the TED problem using group action theory, exploring the connection between tree editing operations and permutation groups. By formulating node insertion, deletion, and relabeling as group actions, we offer a novel perspective on tree transformations. This group-theoretic and metric-based approach provides new insights into the structure of tree similarity and introduces new possibilities for TED applications in various fields.

关键词： data structures Consumer electronics

来源：评论

学校读者我要写书评

暂无评论

Bounding User Contributions in the Worst-Case for User-Level Differentially Private Mean Estimation

arXiv

引用

arXiv 2025年

作者： Rameshwar, V. Arvind Tandon, Anshoo India Urban Data Exchange Program Unit Indian Institute of Science Bengaluru India

In this article, we revisit the well-studied problem of mean estimation under user-level Ε-differential privacy (DP). While user-level Ε-DP mechanisms for mean estimation, which typically bound (or clip) user contributions to reduce sensitivity, are well-known, an analysis of their estimation errors usually assumes that the data samples are independent and identically distributed (i.i.d.), and sometimes also that all participating users contribute the same number of samples (data homogeneity). Our main result is a precise characterization of the worst-case estimation error under general clipping strategies, for heterogeneous data, and as a by-product, the clipping strategy that gives rise to the smallest worst-case error. Interestingly, we show via experimental studies that even for i.i.d. samples, our clipping strategy performs uniformly better that the well-known clipping strategy of Amin et al. (2019), which involves additional, private parameter estimation. © 2025, CC BY.

关键词： Differential privacy

来源：评论

学校读者我要写书评

暂无评论

A General Framework for Clustering and Distribution Matching with Bandit Feedback

引用

IEEE Transactions on Information Theory 2025年第3期71卷 2116-2139页

作者： Yavas, Recep Can Huang, Yuqi Tan, Vincent Y. F. Scarlett, Jonathan CNRS@CREATE DesCartes Program CREATE Tower 138602 Singapore National University of Singapore Department of Computer Science Queenstown119077 Singapore National University of Singapore School of Computing Department of Information Systems and Analytics Queenstown119077 Singapore National University of Singapore Department of Mathematics Department of Electrical and Computer Engineering Queenstown119077 Singapore Institute of Data Science National University of Singapore Department of Computer Science The Department of Mathematics Queenstown119077 Singapore

We develop a general framework for clustering and distribution matching problems with bandit feedback. We consider a K-armed bandit model where some subset of K arms is partitioned into M groups. Within each group, the random variable associated to each arm follows the same distribution on a finite alphabet. At each time step, the decision maker pulls an arm and observes its outcome from the random variable associated to that arm. Subsequent arm pulls depend on the history of arm pulls and their outcomes. The decision maker has no knowledge of the distributions of the arms or the underlying partitions. The task is to devise an online algorithm to learn the underlying partition of arms with the least number of arm pulls on average and with an error probability not exceeding a pre-determined value δ. Several existing problems fall under our general framework, including finding M pairs of arms, odd arm identification, and N-ary clustering of K arms belong to our general framework. We derive a non-asymptotic lower bound on the average number of arm pulls for any online algorithm with an error probability not exceeding δ. Furthermore, we develop a computationally-efficient online algorithm based on the Track-and-Stop method and Frank-Wolfe algorithm, and show that the average number of arm pulls of our algorithm asymptotically matches that of the lower bound. Our refined analysis also uncovers a novel bound on the speed at which the average number of arm pulls of our algorithm converges to the fundamental limit as δ vanishes. © 1963-2012 IEEE.

关键词： Random variables

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：