检索结果-内蒙古大学图书馆

Random epipolar constraint loss functions for supervised optical flow estimation

PATTERN RECOGNITION 2024年 148卷

作者： Fan, Zhengyuan Cai, Zemin Shantou Univ Dept Elect Engn Shantou 515063 Guangdong Peoples R China Key Lab Digital Signal & Image Proc Guangdong Prov Shantou 515063 Guangdong Peoples R China 243 Daxue Rd Shantou 515063 Guangdong Peoples R China

The majority of supervised models estimate optical flow through minimizing the numerical difference between the predicted flow and the ground truth, resulting in the loss of positional and geometric characteristics of the calculated flow fields. In addition, these models require a large number of parameters and high computational cost when computing optical flow. To address these issues, this paper presents a novel loss function and a lightweight framework for optical flow estimation. The proposed loss function, called the random epipolar constraint loss function (RECLoss), incorporates epipolar geometry into supervised optimization to transform the numerical difference into geometry constraint. The RECLoss can make the optical flow estimation models more effective and enhance their generalization abilities. Moreover, the design of RECLoss is more interpretable and the estimated optical flow fields from RECLoss have clearly defined mathematical meanings. A lightweight recurrent neural network for optical flow estimation (LRFlow) that balances computational cost and estimation accuracy is also proposed. The LRFlow, containing only 3.0M parameters, consists of a feature extractor, a correlation matching module, and an iterative update unit. The proposed lightweight network achieves state-of-the-art results compared to all other lightweight networks on the challenging MPI-Sintel and KITTI2015 datasets. The effectiveness of RECLoss in improving the accuracy of LRFlow and other state-of-the-art methods such as RAFT and GMA has been validated through extensive experiments. The source code of the project are available at https://***/Eryo-iPython/RECLoss.

关键词： optical flow estimation Epipolar geometry Lightweight neural network

来源：评论

学校读者我要写书评

暂无评论

Ocean Litter detection using Inception Transfer Learning models 15

Ocean Litter detection using Inception Transfer Learning Mod...

引用

Ocean Sensing and Monitoring XV 2023

作者： Sanjai, P. Bonny, Talal Nasir, Nida AlShabi, Mohammad Al Shammaa, Ahmed University of Sharjah United Arab Emirates College of Computing and Informatics University of Sharjah Sharjah United Arab Emirates Mechanical and Nuclear Department College of Engineering University of Sharjah Sharjah United Arab Emirates Department of ECE Karunya Institute of Technology and Sciences Coimbatore India University of Khorfakkan Sharjah United Arab Emirates

ISBN: (纸本)9781510662025

Plastic pollution has emerged as one of the biggest environmentally threatening issues. Using image classification, the proposed study aids in categorizing the level of marine pollution in ocean underwater regions. This study classified the amount of pollution in the ocean using the two variants of Inception Convolutional neural network (CNN) models i.e., Inception- ResNet V2, and InceptionV3. High accuracies of up to 96.4% have been reported. This study will help researchers working in the field of water quality detection. © 2023 SPIE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Image classification and discrimination of COVID pneumonia based on convolutional neural network 7

Image classification and discrimination of COVID pneumonia b...

引用

7th International Conference on Mechatronics and Intelligent Robotics, ICMIR 2023

作者： Qin, Chuan Liu, Xuesong Che, Yuanxi Yao, Enjie School of Computing Newcastle University Newcastle Upon TyneNE1 7RU United Kingdom Department of Electronic and Electrical Engineering The University of Sheffield SheffieldS10 2TN United Kingdom Department of Computer Science and Technology Northwest University Xi an710127 China

ISBN: (数字)9781510668089

ISBN: (纸本)9781510668072

To address the negative quality of RT-PCR and the time taken to obtain results, this paper reports on the classification and discrimination of COVID pneumonia images by developing a neural network adaptively. Given the limited number of published COVID-19 CT data currently available, we suggest using the proposed negative process to refine the data to obtain more CT data to reduce matching risk;Experimental results show that compared to network models such as AlexNet and GoogleNet, the proposed BUF-Net network model has the best performance with an accuracy of 93%. Visualization of systemic findings using Grad-CAM techniques may further elucidate the critical role of CT imaging in the diagnosis of ***: The use of deep learning tools in clinical practice can help radiologists make better diagnoses. © 2023 SPIE. All rights reserved.

关键词： Coronavirus

来源：评论

学校读者我要写书评

暂无评论

Lightweight network Research based on deep learning

Lightweight Network Research based on deep learning

引用

2021 International Conference on Computer Graphics, Artificial Intelligence, and Data Processing, ICCAID 2021

作者： Zheng, Meijun Tian, Yimin Chen, Hongmei Yang, Shuai Song, Fangfang Gao, Xue School of Information Engineering Beijing Institute of Graphic Communication Beijing China

ISBN: (数字)9781510652170

ISBN: (纸本)9781510652163

Convolutional neural network models in the field of computer vision are also emerging in endlessly. From LeNet in 1998, to AlexNet, which detonated the deep learning boom in 2012, to VGG[1] in 2014, and ResNet[2] in 2015, the application of deep learning network models in image processing is getting better and better. neural networks are getting larger and larger, their structures are becoming more and more complex, and the hardware resources required for prediction and training are gradually increasing. Often, deep learning neural network models can only be run on servers with high computing power. Deep convolutional neural networks have improved the performance of multiple computer vision tasks to a new level. The overall trend is to build deeper and more complex networks in order to achieve higher accuracy, but these networks may not necessarily satisfy mobile devices in terms of scale and speed. Requirements. Due to the limitations of hardware resources and computing power, it is difficult for mobile devices to run complex deep learning network models. Efforts are also being made in the field of deep learning to promote the miniaturization of neural networks. While ensuring the accuracy of the model, it is smaller and faster. From 2016 until now, the industry has proposed lightweight network models such as SqueezeNet[3], ShuffleNet V1-V2[4-5], MobileNet V1-V3[6-8], and InceptionV1-V4[9-12]. These models make it possible for mobile terminals and embedded devices to run neural network models. Therefore, this article compares the mainstream lightweight neural network models, analyzes and contrasts various networks, and looks forward to the future development trend of lightweight neural networks and several directions that can be studied. © 2022 SPIE.

关键词： neural network models

来源：评论

学校读者我要写书评

暂无评论

Reference-based Video Colorization with AB chrominance point and Temporal Propagation 24

Reference-based Video Colorization with AB chrominance point...

引用

16th International Conference on Machine Learning and computing (ICMLC)

作者： Yang, Yizhang Jin, Jianxiu Huang, Yin Guo, Kailing Xu, Xiangmin South China Univ Technol Guangzhou Peoples R China Pazhou Lab Guangzhou Peoples R China

ISBN: (纸本)9798400709234

Deep learning has achieved remarkable success in video colorization problems. Reference-based video coloring problem is an important topic. In this paper, we propose a reference-based video colorization model that distinguishes itself from conventional colorization approaches. Unlike exemplar-based colorization methods that focus on leveraging color information from the reference frame, our model addresses the inherent limitation of the reference frame potentially lacking essential color details. To tackle this issue, we introduce the AB chrominance point to supplement the missing color information in the reference frame, yielding more precise colorization results. While traditional models often employ one-way or simple bidirectional optical flow, we use a bidirectional optical flow propagation network to make full use of the information from the warping images and increase the temporal consistency. Experimental results substantiate the efficacy of our approach, showcasing superior performance compared to state-of-the-art methods on benchmark datasets.

关键词： neural networks video colorization exemplar-based temporal consistency

来源：评论

学校读者我要写书评

暂无评论

Hybrid learning model for spatio-temporal forecasting of PM2.5 using aerosol optical depth

引用

neural computing & APPLICATIONS 2022年第23期34卷 21367-21386页

作者： Nath, Pritthijit Roy, Biparnak Saha, Pratik Middya, Asif Iqbal Roy, Sarbani Jadavpur Univ Dept Comp Sci & Engn Kolkata India Jadavpur Univ Dept Instrumentat & Elect Engn Kolkata India SRM Univ Dept Comp Sci Chennai Tamil Nadu India

Existence of several challenges and high cost in the development of monitoring infrastructure have become major reasons for data sparsity by statutory government agencies tasked to study pollution exposure in urban areas. As an effort to mitigate this problem, the recent usage of satellite aerosol optical depth data along with the usage of learning algorithms have become popular in recent times. This paper presents a novel four-staged approach using different machine learning, deep learning and statistical methods to develop a spatio-temporal hybrid model for temporal forecasting using data from existing stations along with satellite aerosol optical depth data for spatial interpolation. Experiments conducted on real-world data belonging to the cities of Kolkata, Bengaluru and Mumbai show that a consistent pattern is not followed in all the cities in all stages except in spatial interpolation where Random Forest Regression is found to surpass all other models used. While a long short-term memory network (LSTM Auto-Encoder) when employed in temporal forecasting inside the hybrid method outperforms others in Mumbai, a random forest regression-based method and a multi-layer perceptron-based method outperform others similarly in Kolkata and Bengaluru, respectively.

关键词： Air pollution Hybrid models Aerosol optical depth Spatio-temporal modelling

来源：评论

学校读者我要写书评

暂无评论

Photonic Bayesian neural network Using Programmed optical Noises

引用

IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS 2023年第2期29卷

作者： Wu, Changming Yang, Xiaoxuan Chen, Yiran Li, Mo Univ Washington Dept Elect & Comp Engn Seattle WA 98195 USA Duke Univ Dept Elect & Comp Engn Durham NC 27708 USA Univ Washington Dept Phys Seattle WA 98195 USA

The Bayesian neural network (BNN) combines the strengths of neural networks and statistical modeling in that it simultaneously performs posterior predictions and quantifies the uncertainty of the predictions. Integrated photonics has emerged as a promising hardware platform of neural network accelerators capable of energy-efficient, low latency, and parallel computing. However, photonic neural networks demonstrated to date are mostly deterministic network models. Here, we extend the photonic neural network to a statistical model and proposed a photonic Bayesian neural network (P-BNN) architecture based on the integrated photonic platform and harnessing the inherent optical noises. The Bayesian neuron is realized by controlling the probability distribution of the signal-amplified spontaneous emission (signal-ASE) beat noise. We show the P-BNN's advantages in making predictions using the posterior distribution by simulating a p-BNN to perform handwritten number classification tasks. The simulation results show that the proposed P-BNN not only makes successful predictions on the expected images from the test dataset but also detects and rejects the unexpected images outside the training datasets. The P-BNN architecture is compatible with on-chip optical amplifiers and can be scaled up using current and emerging integrated photonics technologies, thus is promising for practical neural network applications.

关键词： Integrated optics neural network optical computing

来源：评论

学校读者我要写书评

暂无评论

Research and analysis of pruning algorithm

Research and analysis of pruning algorithm

引用

2022 International Conference on Algorithms, Microchips and network Applications

作者： Jin, Xin Ma, Xuanmin Tao, Li College of Communication and Information Technology Nanjing University of Posts and Telecommunications Nanjing210023 China Leicester International Institute Dalian University of Technology Panjin124221 China College of Computer Science Sichuan University Chengdu610207 China

ISBN: (纸本)9781510653290

In recent years, thanks to the development of GPU (Graphic Processing Unit) computing power, deep neural network (DNN) has drawn extensive attention from academia and industry and has been widely used in various fields of artificial intelligence. However, the mainstream deep learning models have some problems such as large storage space consumption and high computational complexity, which makes it difficult for the deep neural network model to be deployed to mobile devices with limited computing resources or applications with strict delay requirements. Therefore, on the premise of minimizing the loss of model accuracy, the algorithm researches of deep neural network compression and acceleration which can lighten the model has been carried out. In recent years, model compression methods such as pruning, quantification and knowledge distillation have emerged. This paper focuses on neural network pruning algorithm. Firstly, this paper expounds the basic methods and classification of neural network pruning algorithms. Then according to the classification of pruning granularity, a more detailed classification, analysis and comparison of specific models are carried out from the algorithm level. Then based on cifar-10 data set, validity tests of some neural network pruning algorithms and contrastive analyses of the experimental data from different models such as the parameters are carried out. Finally, in combination of the recent results of neural network pruning algorithm, we put forward research dilemmas and problems to be solved. Then the future research trends and methods of neural network pruning algorithm are prospected. © COPYRIGHT SPIE. Downloading of the abstract is permitted for personal use only.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Analog Photonics computing for Information Processing, Inference, and Optimization

引用

ADVANCED QUANTUM TECHNOLOGIES 2023年第9期6卷

作者： Stroev, Nikita Berloff, Natalia G. Weizmann Inst Sci Dept Phys Complex Syst IL-76100 Rehovot Israel Univ Cambridge Dept Appl Math & Theoret Phys Cambridge CB3 0WA England

This review presents an overview of the current state-of-the-art in photonics computing, which leverages photons, photons coupled with matter, and optics-related technologies for effective and efficient computational purposes. It covers the history and development of photonics computing and modern analogue computing platforms and architectures, focusing on optimization tasks and neural network implementations. The authors examine special-purpose optimizers, mathematical descriptions of photonics optimizers, and their various interconnections. Disparate applications are discussed, including direct encoding, logistics, finance, phase retrieval, machine learning, neural networks, probabilistic graphical models, and image processing, among many others. The main directions of technological advancement and associated challenges in photonics computing are explored, along with an assessment of its efficiency. Finally, the paper discusses prospects and the field of optical quantum computing, providing insights into the potential applications of this technology.

关键词： coherent Ising and XY machines optical computing optical neural networks optical quantum computing photonics computing

来源：评论

学校读者我要写书评

暂无评论

An Efficient Approach for Pedestrian Detection 15

An Efficient Approach for Pedestrian Detection

引用

15th International Conference on computing Communication and networking Technologies, ICCCNT 2024

作者： Srivastava, Anugrah Badal, Tapas Goyal, Ankush Naik, Advait Scset Bennett University Greater Noida India Institute of Technology Computer Engineering Vivekanand Education Society's Mumbai India

ISBN: (纸本)9798350370249

Pedestrian detection is a significant field within computer vision, closely related to object detection, with applications spanning surveillance, autonomous driving, optical character recognition (OCR), and face recognition. This paper focuses on detecting pedestrians and cyclists using images captured from surveillance cameras and vehicle-mounted cameras. Our application involves detecting objects in complex environments where the distinction between objects and the background is subtle. The detection model must achieve high accuracy and precision. We have employed the Faster R-CNN model for this task. Although Faster R-CNN offers higher accuracy compared to models like YOLO, it operates at a lower speed. The enhanced precision of Faster R-CNN makes it well-suited for our needs, despite the trade-off in processing time. Our implementation includes data preparation, model training with fine-tuning on a custom dataset, evaluation, and calculation of the mean number of bounding boxes. This approach ensures robust detection performance, which is crucial for the intended applications. © 2024 IEEE.

关键词： Convolutional neural network Hyperparameter Tuning Object Detection Region Proposal

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：