Xing Lihua
When conducting high-speed railway tunnel construction, engineering personnel consistently face the operation of risk factors in varying environments due to constraints imposed by geographical conditions and related technical means. In recent years, a series of safety accidents have frequently occur…
Zhao Qi
With the rapid development of society, the transformation of educational concepts has emphasized a student-centered approach, with greater attention paid to students' individual development and innovative capabilities. In contemporary teaching space design, relatively more focus is placed on the ext…
Dong Peng
Earth chamber pressure is one of the key parameters for controlling ground settlement. Based on extensive real-time data collected during shield tunneling construction, this paper proposes an earth chamber pressure prediction model for Earth Pressure Balance (EPB) shield construction using the "Rand…
Yan Fang
Based on the close undercrossing of the existing Line 3 by the Jiucheng ventilation shaft—Chengxin Avenue station section of Nanjing Metro Line 5 TA01 bid, this paper introduces a series of targeted measures implemented during shield tunneling through soft strata under the condition of poor deformat…
Huang Yanting
The construction industry is undergoing a historic transformation through digitalization. BIM (Building Information Modeling) technology, with its characteristics of visualization, collaboration, and data-driven approaches, has become a core tool for enterprises to achieve cost reduction and efficie…
Jiang Hu, Wang Meng, Chen Pu, Liu Weiwei
This study investigates red mud tailings material, employing GCTS cyclic triaxial apparatus to conduct dynamic triaxial tests on red mud, examining its dynamic deformation and strength characteristics under varying consolidation confining pressures and consolidation ratios. Results indicate that wit…
Lu Beibei
Design management in EPC renovation projects not only serves as the driving force at the project's initiation but also permeates throughout the entire project life cycle. This paper summarizes design management strategies for EPC renovation projects by citing actual engineering cases. Through compre…
Pan Haiping
As the proportion of aging population in our country accelerates, the silver economy emerges, making demands for medical care, elderly care, and social services critical issues that society should actively address. This study aims to explore pathways for enhancing the quality of life and environment…
Gu Yi
With the continuous deepening of globalization and international exchange, overseas diplomatic architecture is playing an increasingly vital role in showcasing national image and facilitating international engagement. The Vietnam Hanoi Chinese Cultural Center, as a significant overseas project of ou…
Wang Meng
Tailings dams constitute major hazard sources for metal and non-metal mines, and red mud represents a strongly alkaline tailing discharged during alumina extraction processes. In recent years, the number of tailings dams has been continuously increasing, with earthquakes serving as critical factors …
Luriel, Liu Zhijun, Sun Guanhua, Haokun Shen, Lu Yingfa
Based on the characteristics of small deformation theory, a new method for stability analysis of slope retaining walls under different rotation axis conditions is proposed (small deformation theory solution). For the slope body: based on the stress distribution satisfying the stress equilibrium diff…
Tang Aisong, Wang Yongsheng, Yao Shaoqiang
As a pyroclastic rock, tuff finds extensive application in construction and engineering due to its dense structure, high compressive strength, and excellent durability. This study investigates the in-situ mechanical properties of tuff from sites in Shunxi (Zhejiang), Fushun (Liaoning), and Emin (Xin…
Liu Jun
To provide appropriate and accurate shear strength parameters for the strongly weathered granite at a nuclear power plant site, field shear tests and laboratory simulated shear tests were conducted based on analysis of its fundamental characteristics. The study concludes that: the cohesion and frict…
Dang Shengshuo, Ding Wenqi, Zhang Qingzhao, Wang Qiushi, Tang Jixiang
Taking the water inrush event in the Wangjiazhai semi-lithified tunnel in Yunnan as the engineering background, this study analyzes the video characteristics throughout the entire water inrush process, selects a deep learning model suitable for identifying tunnel water inrush features, and utilizes …
Cui Zhuoyu, Ding Wenqi, Shen Yude, Tang Jixiang, Ma Chang
Segment assembly, as a crucial procedure in shield tunneling construction, renders segment layout technology of significant importance for tunneling projects. Currently, layout techniques relying on designers' experience and traditional mathematical formulas exhibit low precision and efficiency. To …
Du Weizhang, Cheng Peiming, Zhao Shunli, Yang Yu, Guo Chong
To address the issues of harsh working environments, low efficiency for technicians, and insufficient adaptability to complex geological conditions in traditional rock mass in-situ test specimen preparation, this paper proposes an in-situ specimen preparation device based on high-pressure water-abra…
Xuefen Han, Shen Fei, Zhou Bin, Cai Xiaoxiao, Yi Tiancheng, Hu Zhiliang, Wang Songlin, Liang Tianjiao, Robert Golub, Liang Tianjiao
The ultracold neutron (UCN) transport code, MCUCN, designed initially for simulating UCN transportation from a solid deuterium ($\rm SD_{2}$) source and neutron electric dipole moment experiments, could not simulate UCN storage and transportation in a superfluid $\rm{^{4}{He}}$ (SFHe, He II) source …
Liu Yihong, Xiao Yunzhi, Yu Lingda, Yang Xing, Ma Xiaochao, Leng Yongbin
To monitor and analyze the beam loss characteristics during Hefei Light Source operation, a high-time-resolution synchronous monitoring system for beam loss and bunch position was developed based on a scintillator detector, strip-electrode beam position monitor (Beam Position Monitor, BPM), and high…
Li, Mr. Jiancong, wang, Dr. Shengxiang, Shu, Mr. Xin, Dong, Le, Dr. Yong Lei, Chen, Prof. Jie, wang, Dr. Shengxiang
Neutron CT imaging offers unique advantages in metal defect detection and cultural relic analysis, particularly for exploring internal structures of ancient artifacts like writing knives, due to its high penetrability and hydrogen sensitivity. Accurate segmentation of its images is critical for defe…
Shu, Xin, wang, Dr. Shengxiang, Chen, Prof. Jie, Dr. Xin Ge, Li, Jiancong, Dong, Le, wang, Dr. Shengxiang
X-ray and neutron computed tomography (CT) are widely used for non-destructive characterization of internal structures. Their distinct contrast mechanisms, arised from different interactions with matter, can enhance structural analysis when combined. Dual-modality imaging integrating both techniques…
Peng Wang, Yan Ping, Zhou Yaping, Xiang Yanhui, Xiang Yanhui
Does marriage alter the developmental trajectory of individuals' life satisfaction? This is both a major issue concerning national harmony and stability, and an important question regarding individual quality of life. Based on hedonic adaptation theory and the buffer model, this study utilizes six-w…
Cui,Xuetao, Wu,Qigang, Wang, Jian, Lei,Mingzhun, Song,Yuntao, Cui, Xuetao
Objective: As the primary functional component of a fusion reactor, the fusion blanket pebble bed, composed of numerous particles, is crucial for tritium breeding, neutron multiplication, and radiation shielding. Particles within tritium-breeding pebble beds are subjected to prolonged neutron irradi…
Shan Chuchao
Abstract: As the application of artificial intelligence technology continues to deepen, smart library services are undergoing an intelligent transformation from traditional models, with increasing emphasis on personalized services to deliver high-quality user experiences. Currently, following the di…
Liu Zhifang, Feng Ximei, \documentclass[twocolumn]{article}
\usepackage[utf8]{inputenc}
\usepackage{amsmath}
\usepackage{graphicx}
\usepackage{CJK}
\usepackage[numbers]{natbib}
\title{Research on Image Classification Methods Based on Deep Learning}
\author{Zhang San \\ School of Computer Science and Technology \\ Peking University}
\date{\today}
\begin{document}
\begin{CJK}{UTF8}{gbsn}
\maketitle
\begin{abstract}
This paper studies image classification methods based on deep learning. It first introduces the fundamental principles and commonly used models of deep learning, and then focuses on analyzing the application of convolutional neural networks in image classification tasks. Experimental results demonstrate that the proposed method achieves 92.5\% accuracy on the CIFAR-10 dataset, outperforming traditional machine learning methods.
\end{abstract}
\keywords{deep learning, image classification, convolutional neural network}
\section{Introduction}
With the rapid development of artificial intelligence technology, machine learning has become an important research direction in the field of computer science. In particular, breakthroughs in deep learning technology in recent years have led to significant progress in tasks such as image classification, speech recognition, and natural language processing. This paper will focus on exploring the application of deep learning in image classification.
\section{Related Work}
Traditional image classification methods mainly rely on hand-crafted feature extractors, such as SIFT\cite{lowe2004distinctive} and HOG\cite{dalal2005histograms}. These methods require domain expert knowledge and have limited generalization capabilities. In contrast, deep learning methods can automatically learn hierarchical feature representations, significantly improving classification performance.
\section{Deep Learning Methods}
\subsection{Convolutional Neural Networks}
Convolutional Neural Networks (CNN) are a type of deep learning model specifically designed for processing data with grid-like structures. They extract image features through a combination of convolutional layers, pooling layers, and fully connected layers. Convolutional layers perform convolution operations on the input using learnable filters, while pooling layers are used to reduce feature dimensions.
\subsection{Residual Networks}
Residual Networks (ResNet) solve the vanishing gradient problem in training deep networks by introducing skip connections. This structure enables the network to learn residual mappings, thereby allowing the training of deeper network architectures. ResNet achieved breakthrough results in the ImageNet Large Scale Visual Recognition Challenge.
\section{Experimental Setup}
We evaluated the proposed method on the CIFAR-10 dataset. This dataset contains 60,000 32×32 color images divided into 10 categories. We used 50,000 images for training and 10,000 images for testing. The experiments were implemented using the PyTorch framework and trained on an NVIDIA Tesla V100 GPU.
\section{Experimental Results}
As shown in Table\ref{tab:results}, our method achieved 92.5\% accuracy on the test set. Figure\ref{fig:confusion_matrix} shows the confusion matrix, which demonstrates that the model performs relatively balanced across all categories. Compared with the baseline method, the accuracy improved by 5.2 percentage points.
\begin{table}[h]
\centering
\caption{Performance Comparison of Different Methods}
\label{tab:results}
\begin{tabular}{lc}
\hline
Method & Accuracy(\%) \\
\hline
Traditional machine learning method & 87.3 \\
Deep learning method & 92.5 \\
\hline
\end{tabular}
\end{table}
\section{Conclusion}
This paper proposes an image classification method based on deep learning and validates its effectiveness on the CIFAR-10 dataset. Experimental results demonstrate that deep learning methods have significant advantages in image classification tasks. Future work will explore more efficient model architectures and training strategies.
\bibliographystyle{plain}
\bibliography{references}
\end{CJK}
\end{document}, Liu Zhifang
This study utilized the fixation-related potential technique to examine predictive processing in Chinese reading and the cognitive mechanisms underlying its influence on lexical recognition effects. Analyses of eye movement and electroencephalographic (EEG) data in the pre-target and target word reg…
Tian, Dr. Long, Gao, Dr. Han, Prof. Tianhao Wang, Teng, Mr. Haiyun, tang, Dr. jian, Zheng Qingbo, Zuo, Dr. Taisen, Tengfei Cui, Ph.D., Bin Wang, xu, Mr. Qin, Qiu, Mr. Yongxiang, Yuchen Dong, Zheng, Yujie, Qin, Dr. Zecong, Han, Dr. Zehua, Zhang, Dr. Junpei 张俊佩, Cheng, Dr. He, Prof. Xin Tong (童欣), Prof. Xin Tong
Polarization-analyzed small-angle neutron scattering (PASANS) is an advanced technique that enables the selective investigation of magnetic scattering phenomena in magnetic materials and distinguishing coherent scattering obscured by an incoherent background, making it particularly valuable for cutt…
Zhang Yongqing, Li Na, Gao Yili, Qin Jiawen, Yu Haiping, Shi Hui, Shi Hui
Background Hearing loss is a common health issue among middle-aged and elderly individuals, affecting quality of life and social interaction, and is associated with the risk of cognitive decline. In recent years, population aging in China has intensified, with the proportion of middle-aged and elder…
Zhang Hanzhi, Jinhua, Ma Le, Shi Ling, Chen Chen, Huan Hongmei, Yu Dehua, Yu Dehua
Background The diagnostic and therapeutic capacity for osteoporosis in Shanghai's community health service institutions remains unclear, lacking systematic evaluation and monitoring. Objective To understand the current status of osteoporosis diagnosis, treatment, and management in Shanghai's communi…
Liqing Zhang, Liqing Zhang
Single crystal GaN epilayers were irradiated with heavy inert-gas ions (2.3-MeV Ne8+, 5.3-MeV Kr19+) to fluences ranging from 1.0×1011 to 1.0×1015 ions/cm2. The strain-related damage accumulation versus ion fluences was studied by using a high-resolution X-ray diffractometer (HRXRD) and an ultraviol…
Nian Jingqing, Tan Jiang, Zhang Di, Zhang Yu, Luo Yu, Luo Yu
Threat stimuli possess strong priority in attentional processing, yet whether they facilitate or interfere with the formation of learned distractor inhibition remains unclear. This study examined the influence of threat stimuli on the formation of location-probability-based learned distractor inhibi…
Xin-Yu Yang, Yang, Dr. Jian, Zeng, Dr. GQ, Deng, Mr. Haowen, Tian, Chengshuai, HU, Dr. Chuan-hao, Yang, Dr. Jian
Abstract: For radioactive source localization, traditional Compton cameras are limited by the imaging field of view and low detection efficiency, resulting in slow response in complex environments. To address the limitations in the azimuthal sensitivity of conventional dual-layer scintillator detect…
Mr. Jiaxuan Luo, Li, Mr. Xinbai, Dr. Zebo Tang, Wu, Xin, Prof. Shuai Yang, Zha, Dr. Wangmei, Zhang, Zhan, Dr. Wangmei Zha
The initial collision geometry, including the reaction plane, is crucial for interpreting collective phenomena in relativistic heavy-ion collisions, yet it remains experimentally inaccessible through conventional measurements. Recent studies propose utilizing photon-induced processes as a direct pro…
Mr. Huibin Pi, Dr. Longsheng Wu, Luo, Dr. Deng, Yu, Dr. Guofang, Zheng, Qiwen, Wang, Dr. Xin, Liu, Dr. Xiaonian, Dr. Wangyong Chen, Dr. Longsheng Wu
This study investigates the synergistic effects of cryogenic temperatures and radiation exposure on aerospace components. Test chips were fabricated using a sub-20 nm bulk silicon FinFET process, primarily comprising ring oscillators and bandgap reference (BGR) circuits that incorporate parasitic ve…
Hanxin Liu, Yisheng Song, Yisheng Song
The image noise level indicates the degree to which an image is contaminated by noise, serving as a crucial parameter in image processing. For the third-order tensor corresponding to a color image, traditional methods for estimating the noise level disrupt the data structure of the tensor. To addres…
Wen, Mr. Shihai, Mr. She Qianshun, Kong, Dr. Jie, Dr. Kailong Wang, An, Yilang, Yang, Mr. Zuoqiao, Dr. Jun-Wei Yan, Qian, Dr. Yi, Yu, Dr. Yuhong, Dr. Guangshun Li, Prof. Zhiyu Sun, Dr. Jie Kong
In the High Energy FRagment Separator (HFRS) at the High Intensity heavy-ion Accelerator Facility, plastic scintillation detectors are used for high-precision time-of-flight (ToF) measurements to identify the secondary ions from the reaction between the primary beam and the target. During the day-on…
Zeng Rongtian, Shan Jianqiang, Ge Li, Wang Jianwei, Wu Pan
To investigate the safety and reliability of space nuclear power systems, a transient safety analysis code NUSOL-LMR-Li was developed for the lithium-cooled fast reactor Stirling thermoelectric conversion system, focusing on analyzing its inherent safety under accident conditions and the adaptive ch…
Xu Chenyu, Hu Yingguo, Li Peng, Liu Meishan, Li Gengquan
The Finite-Discrete Element Method (FDEM) is an advanced numerical computational method highly suitable for simulating the complete process of rock blasting and fracture. Considering that intact rock masses contain numerous joint surfaces, this study introduces a realistic joint constitutive model t…
Zhang Boyi, Zhao Chengqianyuan, Sun Zeyuan, Xing Jinquan, Luo Yinfeng, Luo Qinyuan
Rockburst, as a serious geological hazard, frequently poses significant threats to tunnel stability and construction safety. Traditional circular tunnels under non-hydrostatic pressure conditions readily form localized stress concentration zones in the surrounding rock mass, which can easily trigger…
Erfeng Zhang
Soil arching effect; Expanding arch; Triangular slip surface; Limit equilibrium; Tunnel earth pressure
Zhang Minrui, Kong Deheng, Xiao Feizhi, Li Bo
Rockfall disasters along highway slopes in China are characterized by wide distribution, high frequency, and sudden occurrence, presenting a challenge for prevention and control efforts due to the lack of reliable early warning indicators. This study innovatively proposes an electromagnetically cont…
Zhou Qinglong, Lin Wancang
Currently, high-performance computing for large-scale finite element methods primarily relies on parallel computing with multi-CPU computer clusters, which incurs high computational costs and offers limited improvements in computational efficiency. Based on the advantages of current state-of-the-art…
Xiao-Qi Han, Ze-Feng Gao, Zhong-Yi Lu, Peng-Jie Guo, Ze-Feng Gao, Zhong-Yi Lu
The discovery of physical laws has traditionally relied on human intuition, analytical reasoning, and experimental observation. However, modern physics research is increasingly constrained by numerous challenges, such as high specialization, fragmented workflows, and limited computational resources,…
Hu Jiaqi, Shang Peng, Zhu Jinhui, Zuo Yinghong, Liu Li, Niu Shengli, Wang Xuedong
Severe nuclear accidents (such as severe reactor accidents and terrorist attacks involving radioactive materials) are typically accompanied by violent explosions, and the air surrounding the leaked fission products will inevitably be disturbed by shock waves. To reasonably assess the delayed nuclear…
Zhu Haoyuan, Wang Huacai
Mobile hot cell is a radioactive material handling facility that can be transported integrally and rapidly reassembled at the target site, which addresses the issues of poor flexibility and mobility inherent in traditional fixed hot cells through modular design. With the development of the nuclear i…
Dr. Haipeng Zhi, Prof. Jiaru Shi (Accelerator), Zha, Dr. Hao, Dr. Jian Gao, Gao, Dr. Qiang, Chen, Dr. Huaibi, Shi (Accelerator), Prof. Jiaru Shi Jiaru
In this study, a compact X-band dual-mode electron linear accelerator (linac) was developed for medical radiotherapy applications, which is capable of generating both 100 mA / 6 MeV beams for X-ray radiotherapy (low-energy mode) and 10 mA / 13.5 MeV beams for electron radiotherapy (high-energy mode)…
Zhang Lei
To thoroughly investigate the degree of surrounding rock damage during the bench blasting process in tunnel engineering, this study takes a certain Grade II surrounding rock as an example, starting from the mechanism of crack generation in a single blasthole, exploring the interaction mechanism betw…
Yin Yongjin
The Juji Tunnel of the Lida Expressway traverses complex and variable highly weathered, thin-bedded shale strata. The surrounding rock is inherently weak and fractured, with severe large deformation occurring after tunnel excavation. The overall stability of the surrounding rock is poor, and under t…
Tian Xiaolu
To investigate the seepage behavior in karst tunnel surrounding rock, the Yesanguan Tunnel of the Yi-Wan Railway was selected as the engineering case study. Key factors influencing surrounding rock seepage were identified through analysis of engineering data and literature review. An elastic damage-…
Wang Yuedong
Karst geological regions, due to their unique geological structures and surface morphologies, frequently encounter hydrogeological issues such as water inrush, posing potential threats to the stability and safety of urban tunnels. This study investigates the impact of karst geological water inrush o…
Lu Jiafu
Sleeve valve pipe grouting technology has been employed for many years with numerous engineering applications. However, Guiyang Metro planning was initiated relatively late. To study the construction technical parameters of sleeve valve pipe grouting under the geological conditions of Guiyang, this …
Zhang Shengli
The bench method exhibits significant advantages in saving construction time and improving mechanized operation levels, but has certain limitations in controlling surrounding rock stability and preventing deformation. To reduce the deformation impact of bench excavation on tunnel surrounding rock, t…