检索结果-内蒙古大学图书馆

4th International Conference on Intelligent Systems and Pattern Recognition-ISPR

作者： Sehairi, Kamal Bouhafs, Abdelkader Boucherit, Ibtissam Mama Bouwmans, Thierry Chouireb, Fatima Ecole Normale Super Laghouat Dept Phys Lab Appl Sci & Didact Laghouat Algeria Univ Amar Telidji Laghouat Dept Elect Telecommun Signals & Syst Lab Laghouat Algeria Univ La Rochelle Dept Comp Sci Lab Math Image & Applicat La Rochelle France

ISBN: (纸本)9783031821523;9783031821530

Identifying and locating objects in images and videos, including elements like traffic signs, vehicles, buildings, and people, constitutes a fundamental and demanding task in computer vision, known as object detection. Due to the higher computing complexity of this technique and the large amount of data carried by the video signal, it is nearly impossible for ordinary general-purpose processors GPPs or CPUs to run these techniques in real-time, especially for embedded systems applications. Therefore, special hardware that can acquire, control, or execute in parallel is required. These specialized hardware systems include Digital Signal Processors DSPs, Field Programmable Gate Arrays FPGAs, visual processing Units VPUs, Tensor processing Units TPUs, Neural processing Units NPUs or Graphics processing Units GPUs. This work presents the benefits of accelerating traditional object detection methods on a high-end embedded system, the Jetson Nano Developer Kit. This single computer board is equipped with the Tegra K1 System on Chip SoC, which is composed of a quad-core ARM A15 and 192 cores of Kepler-embedded GPU. Computing acceleration was ensured via the use of the CUDA OpenCV library for both the Histogram of Oriented Gradients HOG and the Haar Cascade Classifier. For VGA resolution, results reveal that the GPU implementation on this embedded system is 1.4x faster than the CPU for the HOG method and 2x for the Haar Cascade Classifier method.

关键词： Object detection Machine Learning Haar Cascade HOG Local feature descriptors GPU implementation OpenCV CUDA SoC Tegra X1 Jetson Nano

来源：评论

学校读者我要写书评

暂无评论

Deep Learning-Based Blind image Super-Resolution using Iterative Networks

Deep Learning-Based Blind Image Super-Resolution using Itera...

引用

IEEE International Conference on visual communications and image processing (VCIP) - visual communications in the Era of AI and Limited Resources

作者： Yaar, Asfand Ates, Hasan F. Gunturk, Bahadir K. Istanbul Medipol Univ Sch Engn & Nat Sci Istanbul Turkey

ISBN: (纸本)9781728185514

Deep learning-based single image super-resolution (SR) consistently shows superior performance compared to the traditional SR methods. However, most of these methods assume that the blur kernel used to generate the low-resolution (LR) image is known and fixed (e.g. bicubic). Since blur kernels involved in real-life scenarios are complex and unknown, performance of these SR methods is greatly reduced for real blurry images. Reconstruction of high-resolution (HR) images from randomly blurred and noisy LR images remains a challenging task. Typical blind SR approaches involve two sequential stages: i) kernel estimation;ii) SR image reconstruction based on estimated kernel. However, due to the ill-posed nature of this problem, an iterative refinement could be beneficial for both kernel and SR image estimate. With this observation, in this paper, we propose an image SR method based on deep learning with iterative kernel estimation and image reconstruction. Simulation results show that the proposed method outperforms state-of-the-art in blind image SR and produces visually superior results as well.

关键词： Super-resolution Kernel estimation Blind Iterative Deep learning

来源：评论

学校读者我要写书评

暂无评论

Are Standard CNNs Good Enough for No-Reference Stereoscopic image Quality Assessment?

Are Standard CNNs Good Enough for No-Reference Stereoscopic ...

引用

International Conference on Signal processing and communications, SPCOM

作者： Ishita Bardhan Sumohana Channappayya Abhik Banerjee Abdur Rahim Mohammad Forkan Prem Jayaraman Abhinav Kumar Dept. of Artificial Intelligence Indian Institute of Technology Hyderabad India Dept. of Computing Technologies Swinburne University of Technology Melbourne Australia

ISBN: (数字)9798350350456

ISBN: (纸本)9798350350463

Perceptual quality metrics derived from deep features have led to a boost in modelling the Human visual System (HVS) to perceive the quality of visual content. In this work, we study the effectiveness of fine-tuning three standard convolutional neural networks (CNNs) viz. ResNet50, VGG16 and MobileNetV2 to predict the quality of stereoscopic images in the no-reference setting. This work also aims to understand the impact of using disparity maps for quality prediction. Interestingly, our experiments demonstrate that disparity maps do not significantly contribute to improving perceptual quality estimation in the deep learning framework. To the best of our knowledge, this is the first study that explores the impact of disparity along with the chosen models for Stereoscopic image Quality Assessment. We present a detailed study of our experiments with various architectural configurations on the LIVE Phase I and ii datasets. Further, our results demonstrate the innate capability of deep features for quality prediction. Finally, the simple fine-tuning of the models results in solutions that compete with state-of-the-art patch-based stereoscopic image quality assessment methods.

关键词： image quality Measurement Deep learning visualization Stereo image processing Estimation visual systems

来源：评论

学校读者我要写书评

暂无评论

Fidelity-preserving Learning-Based image Compression: Loss Function and Subjective Evaluation Methodology

Fidelity-preserving Learning-Based Image Compression: Loss F...

引用

IEEE visual communications and image processing (VCIP)

作者： Shima Mohammadi Yaojun Wu João Ascenso Instituto Superior Técnico Instituto de Telecomunicações Bytedance

Learning-based image compression methods have emerged as state-of-the-art, showcasing higher performance compared to conventional compression solutions. These data-driven approaches aim to learn the parameters of a neural network model through iterative training on large amounts of data. The optimization process typically involves minimizing the distortion between the decoded and the original ground truth images. This paper focuses on perceptual optimization of learning-based image compression solutions and proposes: i) novel loss function to be used during training and ii) novel subjective test methodology that aims to evaluate the decoded image fidelity. According to experimental results from the subjective test taken with the new methodology, the optimization procedure can enhance image quality for low-rates while offering no advantage for high-rates.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Multimodal Fake News Detection on Fakeddit Dataset Using Transformer-Based Architectures 4th

Multimodal Fake News Detection on Fakeddit Dataset Using Tra...

引用

4th International Conference on Machine Learning, image processing, Network Security and Data Sciences (MIND)

作者： Kalra, Sakshi Kumar, Chitneedi Hemanth Sai Sharma, Yashvardhan Chauhan, Gajendra Singh BITS Pilani Dept CSIS Pilani Campus Pilani 333031 Rajasthan India BITS Pilani Dept HSS Pilani Campus Pilani 333031 Rajasthan India

ISBN: (纸本)9783031243660;9783031243677

Real-time information is transforming due to technological advancements and widespread internet access. In our increasingly digital culture, fake news and misinformation are more common in journalism, news reporting, social media, and other online information consumption platforms. The spread of misinformation can have harmful impacts or even control public events by using multimedia content to deceive readers and gain dissemination. The question here is how to spot fake news about recently occurring events and it is one of the special difficulties in Fake News Detection (FND) on social networking sites. Recent study has considerably increased our ability to identify fake news, because of less emphasis on utilizing the relationship between the textual and visual information in news samples. It is possible to spot fake news by giving importance to similarity among textual and visual features. In this paper, we study the task of identifying fake news using the Fakeddit dataset, which is a collection of full-length articles and related images. We propose a multimodal approach that makes use of transfer learning to gather semantic and contextual data, develop stronger hidden representations between the words in news samples and the images, and tries to improve the accuracy of FND task. We carefully evaluate the performance of our model on the Fakeddit dataset. The results demonstrate that the proposed model learns more accurate textual features and outperforms the most current textual results on that dataset.

关键词： Natural Language processing (NLP) Fake News Detection (FND) Transfer learning Deep learning Social media analytics

来源：评论

学校读者我要写书评

暂无评论

Fine-grained Single-layer Tiling for Viewport-Adaptive 360-degree Video Streaming

Fine-grained Single-layer Tiling for Viewport-Adaptive 360-d...

引用

IEEE visual communications and image processing (VCIP)

作者： Jong-Beom Jeong Jun-Hyeong Park Soonbin Lee Eun-Seok Ryu Department of Computer Science Education Sungkyunkwan University (SKKU) Seoul Republic of Korea Department of Immersive Media Engineering Sungkyunkwan University (SKKU) Seoul Republic of Korea Multimedia Communications Group Fraunhofer Heinrich-Hertz-Institute (HHI) Berlin Germany

Tile-based streaming is widely adopted for viewport-adaptive 360-degree video streaming due to its potential for bitrate reduction. However, for legacy devices equipped with a single decoder, managing multiple tile bitstreams poses a challenge, particularly with regards to synchronizing multiple decoders. This paper introduces Butterfly360, a fine-grained single-layer tiling method that advances two main ideas: (i) an optimal tiling scheme decision method for single-layer merging of tiles using various tile sizes, and (ii) a simple rate adaptation method for viewport tiles. Experimental results demonstrate that the proposed method offers advantages in terms of bitrate savings and reduction in decoding runtime. Furthermore, the proposed method facilitates the use of either multiple tile bitstreams or a single-layer merged bitstream.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Structure-Texture Decomposition-Based Enhancement Framework for Weakly Illuminated images 6th

Structure-Texture Decomposition-Based Enhancement Framework ...

引用

6th International Conference on Computer Vision and image processing, CVIP 2021

作者： Haritha, K.M. Sreeni, K.G. Zacharias, Joseph Jeena, R.S. Department of Electronics and Communication Engineering College of Engineering Trivandrum Thiruvananthapuram India

ISBN: (纸本)9783031113451

images acquired in poor illumination conditions are characterized by low brightness and considerable noise which constrain the performance of computer vision systems. image enhancement thus remains crucial for improving the efficiency of such systems. To improve the visibility of low-light images, a novel image enhancement framework based on the structure-texture decomposition is proposed in this paper. Firstly, the low-light image is split into structure and texture layers using the total-variation (TV) based image decomposition approach. The structure layer is initially diffused using Perona-Malik (PM) diffusion model and the local and global luminance enhancement is incorporated in the structure-pathway using an expanded model of biological normalization for visual adaptation. In the texture pathway, the suppression of local noise and the enhancement of image details are attained with the estimation of the local energy of the texture layer and the strategy of energy weighting. Eventually, the final enhanced image of improved quality is obtained by merging the modified structure and texture layers. The effectiveness of the proposed framework is validated using no-reference image quality metrics including IL-NIQE, BRISQUE, PIQE, and BLiiNDS-ii. The experimental results show that the proposed method outperforms the state-of-the-art approaches. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.

关键词： image enhancement

来源：评论

学校读者我要写书评

暂无评论

Vision Transformer Hashing for image Retrieval

Vision Transformer Hashing for Image Retrieval

引用

2022 IEEE International Conference on Multimedia and Expo, ICME 2022

作者： Dubey, Shiv Ram Singh, Satish Kumar Chu, Wei-Ta Indian Institute of Information Technology Computer Vision and Biometrics Lab Allahabad India National Cheng Kung University Dept. of Computer Science and Information Engineering Taiwan

ISBN: (数字)9781665485630

ISBN: (纸本)9781665485630

Recently, Transformer has emerged as a new architecture in deep learning by utilizing self-attention without convolution. Transformer is also extended to Vision Transformer (ViT) for the visual recognition with a promising performance on imageNet. In this paper, we propose a Vision Transformer Hashing (VTS) for image retrieval. We utilize the pre-trained ViT on imageNet as the backbone network and add the hashing head. The proposed VTS model is fine tuned for hashing under six different image retrieval frameworks with their objective functions. We perform the extensive experiments on CIFAR10, imageNet, NUS-Wide, and COCO datasets. The proposed VTS based image retrieval outperforms the recent state-of-the-art hashing techniques with a significant margin. We also find the proposed VTS model as the backbone network is better than the existing networks, such as AlexNet and ResNet. The code is released at https://***/shivram1987/VisionTransformerHashing. © 2022 IEEE.

关键词： image retrieval

来源：评论

学校读者我要写书评

暂无评论

Pseudo-Supervised Learning for Semantic Multi-Style Transfer

引用

IEEE ACCESS 2021年 9卷 7930-7942页

作者： Kim, Saehun Do, Jeonghyeok Kim, Munchurl Korea Adv Inst Sci & Technol Sch Elect Engn Daejeon 34141 South Korea

Numerous methods for style transfer have been developed using unsupervised learning and gained impressive results. However, optimal style transfer cannot be conducted from a global fashion in certain style domains, mainly when a single target-style domain contains semantic objects that have their own distinct and unique styles, e.g., those objects in the anime-style domain. Previous methods are incongruent because the unsupervised learning can not provide the semantic mappings between the multi-style objects according to their unique styles. Thus, in this paper, we propose a pseudo-supervised learning framework for the semantic multi-style transfer (SMST), which consists of (i) a pseudo ground truth (pGT) generation phase and (ii) a SMST learning phase. In the pGT generation phase, multiple semantic objects of the photo images are separately transferred to the target-domain object styles in an object-oriented fashion. Then the transferred objects are composed back to an image, which is the pGT. In the SMST learning phase, a SMST network (SMSTnet) is trained with the pairs of the photo images and its respective pGT in a supervised manner. From this, our framework can provide the semantic mappings of multi-style objects. Moreover, to embrace the multi-styles of various objects into a single generator, we design the SMSTnet with channel attentions in conjunction with a discriminator dedicated to our pseudo-supervised learning. Our method has been applied and intensively tested for anime-style transfer learning. The experimental results demonstrate the effectiveness of our method and show its superiority compared to the state-of-the-art methods.

关键词： Style transfer image-to-image translation generative adversarial networks

来源：评论

学校读者我要写书评

暂无评论

Improving image Paragraph Captioning with Dual Relations

Improving Image Paragraph Captioning with Dual Relations

引用

2022 IEEE International Conference on Multimedia and Expo, ICME 2022

作者： Liu, Yun Shi, Yihui Feng, Fangxiang Li, Ruifan Ma, Zhanyu Wang, Xiaojie School of Artificial Intelligence Beijing University of Posts and Telecommunications China Beijing Academy of Artificial Intelligence Beijing China

ISBN: (数字)9781665485630

ISBN: (纸本)9781665485630

image paragraph captioning aims to generate multiple de-scriptive sentences for an image. However, most previous methods ignore the explicit relations among objects resulting in unsatisfactory performance. In this paper, we propose a novel model (i.e., DualRel) to capture spatial and seman-tic relations among objects. Specifically, the spatial relation embedding is obtained solely from images using a predefined geometry pattern. With the help of captions, the semantic relation embedding is learned in a weakly supervised man-ner. These two relation embeddings are then interacted with regional features of objects through a relation-aware attention interaction. It first obtains a visual context vector using regional features. Then with the visual context vector, we obtain the corresponding spatial and semantic relation-aware vectors using attentions. These three vectors are fused with two gates for language decoding to further generate a para-graph. Experimental results on Stanford benchmark dataset show that DualRel achieves remarkable improvements11Code released at https://***/fuyunll07/DualRel. © 2022 IEEE.

关键词： Vectors

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：