检索结果-内蒙古大学图书馆

Construction of Deep ReLU Nets for Spatially Sparse Learning

IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING systems 2023年第10期34卷 7746-7760页

作者： Liu, xia Wang, Di Lin, Shao-Bo Xian Univ Technol Sch Sci Xian 710048 Peoples R China Xi An Jiao Tong Univ Ctr Intelligent Decis Making & Machine Learning Sch Management Xian 710049 Peoples R China

Training an interpretable deep net to embody its theoretical advantages is difficult but extremely important in the community of machine learning. In this article, noticing the importance of spatial sparseness in signal and image processing, we develop a constructive approach to generate a deep net to capture the spatial sparseness feature. We conduct both theoretical analysis and numerical verifications to show the power of the constructive approach. Theoretically, we prove that the constructive approach can yield a deep net estimate that achieves the optimal generalization error bounds in the framework of learning theory. Numerically, we show that the constructive approach is essentially better than shallow learning in the sense that it provides better prediction accuracy with less training time.

关键词： Deep learning Neural networks Training Spatial resolution Partitioning algorithms Signal resolution Optimization Constructive deep net (CDN) deep learning learning theory spatial sparseness

来源：评论

学校读者我要写书评

暂无评论

HPDNET: Hyper Prior Dependent Demosaic Neural Network 6

HPDNET: Hyper Prior Dependent Demosaic Neural Network

引用

6th IEEE International conference on image processing, Applications and systems, IPAS 2025

作者： Seo, Youngil Jun, Sungho Lim, Dongpan Song, Seongwook Samsung Electronics System LSI Division Yongin Korea Republic of

ISBN: (纸本)9798331506520

Recent advancements in deep neural networks have shown remarkable improvements in image quality during the demosaicking process, surpassing conventional algorithms. However, these deep neural network techniques are often characterized by heavy computational requirements, rendering them unsuitable for deployment on resource-constrained platforms. This presents a critical challenge in the field of image demosaicking: while deep learning approaches excel in enhancing image quality, their computational intensity poses a significant hindrance to their wider adoption. Consequently, there is a pressing need for methodologies that can strike an optimal balance between achieving superior image quality and maintaining computational efficiency. In this work, we propose a new deep framework, hyper-prior dependent demosaic neural network, HPDNet that utilizes the significant concepts of the conventional algorithm and the characteristic of image sensor data. We designed the network that exploits three concepts, those are pixel gradient prior attention, phase separation, and multi-level sparse and dense feature extraction. We designed the deep neural network that extracts the optimal gradient prior and the multi-level extracted features are fused and attended by gradient prior. It can fully utilize spatially variant information. Also to make the network deployed in the mobile platform, we devised self-pruned image convolution that adopts image filter characteristic and reduce computations. Experiments show that proposed network outperforms SOTA demosaic networks both in terms of image quality and computation. © 2025 IEEE.

关键词： image quality

来源：评论

学校读者我要写书评

暂无评论

Siamese-Driven Optimization for Low-Resolution image Latent Embedding in image Captioning 27

Siamese-Driven Optimization for Low-Resolution Image Latent ...

引用

27th IEEE Signal processing: algorithms, Architectures, Arrangements, and Applications, SPA 2024

作者： Tan, Jing Jie Mokraoui, Anissa Kwan, Ban-Hoe Ng, Danny Wee-Kiat Hum, Yan-Chai Universiti Tunku Abdul Rahman Lee Kong Chian Faculty of Engineering and Science Department of Mechatronics and Biomedical Engineering Malaysia Université Sorbonne Paris Nord Laboratoire de traitement et transport de l'information France

ISBN: (纸本)9788362065486

image captioning is essential in many fields including assisting visually impaired individuals, improving content management systems, and enhancing human-computer interaction. However, a recent challenge in this domain is dealing with low-resolution image (LRI). While performance can be improved by using larger models like transformers for encoding, these models are typically heavyweight, demanding significant computational resources and memory, leading to challenges in retraining. To address this, the proposed SOLI (Siamese-Driven Optimization for Low-Resolution image Latent Embedding in image Captioning) approach presents a solution specifically designed for lightweight, low-resolution images captioning. It employs a Siamese network architecture to optimize latent embeddings, enhancing the efficiency and accuracy of the image-to-text translation process. By focusing on a dual-pathway neural network structure, SOLI minimizes computational overhead without sacrificing performance, making it an ideal choice for training on resource-constrained scenarios. © 2024 Division of Signal processing and Electronic systems, Poznan University of Technology (DSPES PUT).

关键词： Signal encoding

来源：评论

学校读者我要写书评

暂无评论

SEP: Stage-Enhanced Panoptic Segmentation based on Fully Convolutional Networks 15

SEP: Stage-Enhanced Panoptic Segmentation based on Fully Con...

引用

15th International conference on Signal processing systems, ICSPS 2023

作者： Ji, Shucheng Yuan, xiaochen Bao, Junqi Faculty of Applied Science Macao Polytechnic University 999078 China

ISBN: (纸本)9781510675056

Panoptic segmentation is a critical technology in the field of multimedia, applicable to various domains such as autonomous driving and image recognition. However, due to the enormity and complexity of the task, enhancing the efficiency and accuracy of panoptic segmentation remains a challenge. In this paper, we propose a stage-enhanced panoptic segmentation method which improves the feature extraction network of the backbone, incorporates a stage feature fusion network, and designs a module for adaptive stage feature weight allocation. These enhancements optimize the overall network and enrich the stage features. Experimental results on the publicly available COCO-2017 dataset confirm the performance of Stage-Enhanced Panoptic and demonstrate its superiority compared to other methods. © 2024 SPIE.

关键词： image segmentation

来源：评论

学校读者我要写书评

暂无评论

Compact colorful compressive spectral imager based on deep learning reconstruction

Compact colorful compressive spectral imager based on deep l...

引用

2024 International conference on image, Signal processing, and Pattern Recognition, ISPP 2024

作者： Li, Jinshan Ma, Xu Key Laboratory of Photoelectronic Imaging Technology and System of Ministry of Education of China School of Optics and Photonics Beijing Institute of Technology Beijing100081 China

ISBN: (纸本)9781510680425

Leveraging the spatio-spectral modulation and sophisticated reconstruction algorithms, the colorful compressive spectral imaging (CCSI) method can reconstruct a three-dimensional spectral image from a single compressive *** CCSI systems enhance the modulation freedom through the combination of colorful coding mask (CCM) and dispersive element, but this kind of system has complex structure that limits the miniaturization of ***, the reconstruction quality of CCSI systems can be further improved by using deep learning *** paper proposes a compact CCSI method based on deep learning reconstruction, which tries to reduce the volume of system by attaching the CCM to the *** combination of CCM and RGB detector enhances the modulation ***, a Transformer-based deep learning algorithm is used to obtain promising reconstruction results of the target spectral *** of both simulations and experiments demonstrate the effectiveness of the proposed compact CSSI method. © 2024 SPIE.

关键词： image compression

来源：评论

学校读者我要写书评

暂无评论

Joint Dehazing and Low-light image Enhancement with Complete Atmospheric Model 16

Joint Dehazing and Low-light Image Enhancement with Complete...

引用

16th International conference on Signal processing systems, ICSPS 2024

作者： Wu, Yong xia, Kexiang Ge, Liyong Ye, Jiongkai Zhejiang Institute of Communications Hangzhou311112 China The School of Automation Hangzhou Dianzi University Hangzhou310000 China

ISBN: (纸本)9781510689251

Current image dehazing algorithms often encounter issues of contrast reduction and color distortion in the shadow regions of images. To address this challenge, this paper proposes a comprehensive atmospheric model that simultaneously considers haze and low-light conditions. This model incorporates a global darkness factor as a disturbance variable alongside the atmospheric scattering model, enabling adaptive luminance balancing without the need to estimate the global atmospheric light or the global darkness factor. Building on the principle of object color rendition, This paper justifies the use of the dark channel prior for haze removal and extends the bright channel prior for low-light image enhancement. Furthermore, we explore the mapping relationship between normal and inverted images based on the complete atmospheric model. By solving for the transmission rates of the global atmospheric light and global darkness factor, we propose an image enhancement algorithm capable of concurrently processing haze and low-light images. The proposed algorithm effectively mitigates issues of contrast reduction and color distortion. Experimental results demonstrate the efficacy of our approach, showing that the visual outcomes of simultaneous haze removal and low-light enhancement surpass those achieved through single-task haze removal. © 2025 SPIE.

关键词： Demulsification

来源：评论

学校读者我要写书评

暂无评论

MQ-CODER INSPIRED ARITHMETIC CODER FOR SYNTHETIC DNA DATA STORAGE 30

MQ-CODER INSPIRED ARITHMETIC CODER FOR SYNTHETIC DNA DATA ST...

引用

30th IEEE International conference on image processing (ICIP)

作者： Pic, Xavier Dimopoulou, Melpomeni San Antonio, Eva Gil Antonini, Marc Cote Azur Univ I3S Lab Nice France CNRS UMR 7271 Sophia Antipolis France

ISBN: (纸本)9781728198354

Over the past years, the ever-growing trend on data storage demand, more specifically for "cold" data (i.e. rarely accessed), has motivated research for alternative systems of data storage. Because of its biochemical characteristics, synthetic DNA molecules are now considered as serious candidates for this new kind of storage. This paper introduces a novel arithmetic coder for DNA data storage, and presents some results on a lossy JPEG 2000 based image compression method adapted for DNA data storage that uses this novel coder. The DNA coding algorithms presented here have been designed to efficiently compress images, encode them into a quaternary code, and finally store them into synthetic DNA molecules. This work also aims at making the compression models better fit the problematic that we encounter when storing data into DNA, namely the fact that the DNA writing, storing and reading methods are error prone processes. The main take away of this work is our arithmetic coder and it's integration into a performant image codec.

关键词： image compression DNA JPEG 2000 arithmetic coder

来源：评论

学校读者我要写书评

暂无评论

Deepfake image and Video Detection Using Deep Learning algorithms

Deepfake Image and Video Detection Using Deep Learning Algor...

引用

2025 International conference on Multi-Agent systems for Collaborative Intelligence, ICMSCI 2025

作者： Srinivasan, Srivanth Nischitha, P. Chavan, Akshita Mohana Rv College of Engineering® Bengaluru India Rv College of Engineering® Bengaluru India

ISBN: (纸本)9798331509828

Deep learning (DL) algorithms are swiftly finding applications in computer vision and natural language processing. Nonetheless, they can also be employed for creating convincing deepfakes, which are challenging to distinguish from reality. The advancements in image and video technology and tools, especially on social media platforms, potentially lead to misuse for malicious purposes like blackmail or defamation. To tackle this issue, several group of researchers tried upon spreading or creating awareness on real or fake data. The proposed approach involves combining Deepfake generation using GANs and Autoencoders with a Deepfake detection method. The aim of this initiative is exclusively to combat disinformation and online fraud for the welfare of the general population. Deepfakes, products of AI, have become increasingly realistic, rendering it nearly difficult to distinguish the content. Auto-encoders with sufficient time can achieve about 92 % accuracy. As the generator improves, the discriminator performance worsens as it struggles to differentiate real or fake data. A perfect generator results in 50% accuracy. With advancements in computational capacity and data availability, the proposed DDM (Deepfake Detection Model) has achieved greater accuracy rate of up to 92.3%. © 2025 IEEE.

关键词： Adversarial machine learning

来源：评论

学校读者我要写书评

暂无评论

Enhancing Non-Coal Object Recognition Using Deep Learning on Conveyor Belts 24

Enhancing Non-Coal Object Recognition Using Deep Learning on...

引用

7th International conference on Signal processing and Machine Learning, SPML 2024

作者： Li, Yuancheng Hou, Yalei xie, Yehui Sheng, Mingwei Wang, Mei Ma, Tian Xi'an University of Science and Technology Xi'an710054 China State Grid Ningxia Electric Power Limited Company Yinchuan Branch Yinchuan750001 China

ISBN: (纸本)9798400717192

Various non-coal foreign objects in raw coal seriously affect the safety and efficiency of coal production. As an attractive object recognition method, object recognition based on computer vision has achieved fruitful results while the accuracy and speed are still great challenges because of the poor image quality results of high conveyor belt speeding and uneven illumination. In this paper, a non-coal foreign object recognition method based on deep learning was proposed. Firstly, based on the noise reduction, image enhancement and image segmentation, a combinatorial optimization method is proposed to preprocess the low-quality images. Secondly, from three aspects including network structure, activation function and decoupling prediction, an improved YOLOv5 algorithm is proposed. Finally, by fusing the visual attention mechanism, the feature extraction quality of the YOLOv5 model is further optimized to enrich the semantic information of foreign objects in the image. The experimental results show that, the proposed method can obtain 88.14% recognition accuracy for gangue objects and 87.10% recognition accuracy for iron objects. Meanwhile, we have achieved as high as 27fps recognition speed, which can meet the real-time recognition on high-speed coal belt conveyor. CCS CONCEPTS •Computing methodologies →Computer graphics →image manipulation →image processing;•Computing methodologies →Artificial intelligence → Computer vision;•Computer systems organization →Real-time systems;•Computing methodologies → Machine learning. © 2024 ACM.

关键词： Combinatorial optimization

来源：评论

学校读者我要写书评

暂无评论

Hypervolume Maximization: A Geometric View of Pareto Set Learning 37

Hypervolume Maximization: A Geometric View of Pareto Set Lea...

引用

37th conference on Neural Information processing systems (NeurIPS)

作者： Zhang, xiaoyuan Lin, xi Xue, Bo Chen, Yifan Zhang, Qingfu City Univ Hong Kong Dept Comp Sci Hong Kong Peoples R China City Univ Hong Kong Shenzhen Res Inst Hong Kong Peoples R China Hong Kong Baptist Univ Dept Math Hong Kong Peoples R China Hong Kong Baptist Univ Dept Comp Sci Hong Kong Peoples R China

This paper presents a novel approach to multiobjective algorithms aimed at modeling the Pareto set using neural networks. Whereas previous methods mainly focused on identifying a finite number of solutions, our approach allows for the direct modeling of the entire Pareto set. Furthermore, we establish an equivalence between learning the complete Pareto set and maximizing the associated hypervolume, which enables the convergence analysis of hypervolume (as a new metric) for Pareto set learning. Specifically, our new analysis framework reveals the connection between the learned Pareto solution and its representation in a polar coordinate system. We evaluate our proposed approach on various benchmark problems and real-world problems, and the encouraging results make it a potentially viable alternative to existing multiobjective algorithms. Code is available at https://***/xzhang2523/hvpsl/tree/master.

关键词：

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：