检索结果-内蒙古大学图书馆

ieee/CVF conference on computer vision and pattern recognition (cvpr)

作者： Li, Xianhang Wang, Yali Zhou, Zhipeng Qiao, Yu Chinese Acad Sci Shenzhen Inst Adv Technol ShenZhen Key Lab Comp Vis & Pattern Recognit SIAT SenseTime Joint Lab Shenzhen Peoples R China Shenzhen Inst Artificial Intelligence & Robot Soc SIAT Branch Shenzhen Peoples R China

ISBN: (数字)9781728171685

ISBN: (纸本)9781728171685

Temporal convolution has been widely used for video classification. However, it is performed on spatio-temporal contexts in a limited view, which often weakens its capacity of learning video representation. To alleviate this problem, we propose a concise and novel SmallBig network, with the cooperation of small and big views. For the current time step, the small view branch is used to learn the core semantics, while the big view branch is used to capture the contextual semantics. Unlike traditional temporal convolution, the big view branch can provide the small view branch with the most activated video features from a broader 3D receptive field. Via aggregating such big-view contexts, the small view branch can learn more robust and discriminative spatio-temporal representations for video classification. Furthermore, we propose to share convolution in the small and big view branch, which improves model compactness as well as alleviates overfitting. As a result, our SmallBigNet achieves a comparable model size like 2D CNNs, while boosting accuracy like 3D CNNs. We conduct extensive experiments on the large-scale video benchmarks, e.g., Kinetics400, Something-Something VI and V2. Our SmallBig network outperforms a number of recent state-of-the-art approaches, in terms of accuracy and/or efficiency. The codes and models will be available on https://***/xhl-video/SmallBigNet .

关键词： Convolution Three-dimensional displays Two dimensional displays Electron tubes computer vision pattern recognition Semantics

来源：评论

学校读者我要写书评

暂无评论

Joint Learning of Blind Video Denoising and Optical Flow Estimation

Joint Learning of Blind Video Denoising and Optical Flow Est...

引用

ieee/CVF conference on computer vision and pattern recognition (cvpr)

作者： Yu, Songhyun Park, Bumjun Park, Junwoo Jeong, Jechang Hanyang Univ Seoul South Korea Korea Adv Inst Sci & Technol Daejeon South Korea

ISBN: (数字)9781728193601

ISBN: (纸本)9781728193601

Many deep-learning-based image/video denoising models have been developed, and recently, several approaches for training a denoising neural network without using clean images have been proposed. However, Noise2Noise method requires paired noisy data, and obtaining them is occasionally difficult, whereas other existing models trained using unpaired noisy data deliver limited performance. Obtaining an accurate optical flow from noisy videos is also a difficult task because conventional optical flow estimation methods are primarily focused on estimating the optical flow using clean videos. This study proposes a new framework to fine-tune video denoising and optical flow estimation networks using unpaired noisy videos. These two networks are jointly trained to realize synergy;an improvement in the denoising performance increases the accuracy of the flow estimation, and an improvement in the flow-estimation performance enhances the quality of the training data for the denoiser. Our experimental results reveal that proposed approach outperforms the existing training schemes in video denoising and also provides accurate optical flows even when the videos contain a considerable amount of noise.

关键词： Videos Noise reduction Optical flow Noise measurement Estimation Training computer vision

来源：评论

学校读者我要写书评

暂无评论

Mimic The Raw Domain: Accelerating Action recognition in the Compressed Domain

Mimic The Raw Domain: Accelerating Action Recognition in the...

引用

ieee/CVF conference on computer vision and pattern recognition (cvpr)

作者： Battash, Barak Barad, Haim Tang, Hanlin Bleiweiss, Amit Intel Haifa Israel Intel Labs San Francisco CA USA

ISBN: (纸本)9781728193601

Video understanding usually requires expensive computation that prohibits its deployment, yet videos contain significant spatiotemporal redundancy that can be exploited. In particular, operating directly on the motion vectors and residuals in the compressed video domain can significantly accelerate the compute, by not using the raw videos which demand colossal storage capacity. Existing methods approach this task as a multiple modalities problem. In this paper we are approaching the task in a completely different way;we are looking at the data from the compressed stream as a one unit clip and propose that the residual frames can replace the original RGB frames from the raw domain. Furthermore, we are using teacher-student method to aid the network in the compressed domain to mimic the teacher network in the raw domain. We show experiments on three leading datasets (HMDB51, UCF1, and Kinetics) that approach state-of-the-art accuracy on raw video data by using compressed data. Our model MFCD-Net outperforms prior methods in the compressed domain and more importantly, our model has 11X fewer parameters and 3X fewer Flops, dramatically improving the efficiency of video recognition inference. This approach enables applying neural networks exclusively in the compressed domain without compromising accuracy while accelerating performance.

关键词： Digital storage

来源：评论

学校读者我要写书评

暂无评论

Densely Self-guided Wavelet Network for Image Denoising

Densely Self-guided Wavelet Network for Image Denoising

引用

ieee/CVF conference on computer vision and pattern recognition (cvpr)

作者： Liu, Wei Yan, Qiong Zhao, Yuzhi SenseTime Res Hong Kong Peoples R China Harbin Inst Technol Harbin Heilongjiang Peoples R China City Univ Hong Kong Hong Kong Peoples R China

ISBN: (数字)9781728193601

ISBN: (纸本)9781728193601

During the past years, deep convolutional neural networks have achieved impressive success in image denoising. In this paper, we propose a densely self-guided wavelet network (DSWN) for real-world image denoising. The basic structure of DSWN is a top-down self-guidance architecture which is able to efficiently incorporate multi- scale information and extract good local features to recover clean images. Moreover, such a structure requires a smaller number of parameters and enables us to achieve better effectiveness than Unet structure. To avoid information loss and achieve a better receptive field size, we embed wavelet transform into DSWN. In addition, we apply densely residual learning to convolution blocks to enhance the feature extraction capability of the proposed network. At the full resolution level of DSWN, we adopt a double branch structure to generate the final output. One branch of them tends to pay attention to dark areas and the other performs better on bright areas. Such a double branch strategy is able to handle the noise at different exposures. The proposed network is validated by BSD68, Kodak24 and SIDD+ benchmark. Additional experimental results show that the proposed network outperforms most state-of-the-art image denoising solutions.

关键词： computer vision pattern recognition conferences

来源：评论

学校读者我要写书评

暂无评论

An Analytical Framework for Trusted Machine Learning and computer vision Running with Blockchain

An Analytical Framework for Trusted Machine Learning and Com...

引用

ieee/CVF conference on computer vision and pattern recognition (cvpr)

作者： Wang, Tao Du, Maggie Wu, Xinmin He, Taiping SAS Inst Inc Cary NC 27513 USA

ISBN: (数字)9781728193601

ISBN: (纸本)9781728193601

Machine learning algorithms often use data from databases that are mutable;therefore, the data and the results of machine learning cannot be fully trusted. Also, the learning process is often difficult to automate. A unified analytical framework for trusted machine learning has been presented in the literature to address both issues. It is proposed building a trusted machine learning system by using blockchain technology, which can store data in a permanent and immutable way. In addition, smart contracts on blockchain are used to automate the machine learning process. However, in such a blockchain framework, data efficiency is a big concern, because it is very expensive to store a large amount of data on blockchain. On the other hand, machine learning based computer vision systems often rely on a lot of data. Therefore, to fully leverage a blockchain-based machine learning framework for computer vision systems, data efficiency issues must be addressed. This paper investigates how to enhance data efficiency in such a framework to bring computer vision systems to the edge. It presents a three-step approach. First, a lightweight machine learning model is trained on the server layer. Second, the trained model is saved in a special binary data format for data efficiency. Finally, the streaming layer takes these binary data as input and scores incoming new data in an online fashion. Real-time semantic segmentation for autonomous driving is used as an example to demonstrate how this approach works. This paper makes the following contributions. First, it improves the analytical framework for trusted computer vision systems based on blockchain. Second, the real-time semantic segmentation example shows how data-efficient learning for computer vision can be performed on the edge.

关键词： Machine learning computer vision Computational modeling Servers Contracts Automation

来源：评论

学校读者我要写书评

暂无评论

Detecting Deep-Fake Videos from Phoneme-Viseme Mismatches

Detecting Deep-Fake Videos from Phoneme-Viseme Mismatches

引用

ieee/CVF conference on computer vision and pattern recognition (cvpr)

作者： Agarwal, Shruti Farid, Hany Fried, Ohad Agrawala, Maneesh Univ Calif Berkeley Berkeley CA 94720 USA Stanford Univ Stanford CA 94305 USA

ISBN: (纸本)9781728193601

Recent advances in machine learning and computer graphics have made it easier to convincingly manipulate video and audio. These so-called deep-fake videos range from complete full-face synthesis and replacement (face-swap), to complete mouth and audio synthesis and replacement (lip-sync), and partial word-based audio and mouth synthesis and replacement. Detection of deep fakes with only a small spatial and temporal manipulation is particularly challenging. We describe a technique to detect such manipulated videos by exploiting the fact that the dynamics of the mouth shape - visemes - are occasionally inconsistent with a spoken phoneme. We focus on the visemes associated with words having the sound M (mama), B (baba), or P (papa) in which the mouth must completely close in order to pronounce these phonemes. We observe that this is not the case in many deep-fake videos. Such phonemeviseme mismatches can, therefore, be used to detect even spatially small and temporally localized manipulations. We demonstrate the efficacy and robustness of this approach to detect different types of deep-fake videos, including in-the-wild deep fakes.

关键词： computer graphics

来源：评论

学校读者我要写书评

暂无评论

Foreword to the Special Issue on computer vision-Based Approaches for Earth Observation

引用

ieee JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING 2020年 13卷 6515-6518页

作者： Tuia, Devis Hansch, Ronny Yokoya, Naoto Brown, Myron Le Saux, Bertrand EPFL Valais CH-1950 Sion Switzerland German Aerosp Ctr D-82234 Wessling Germany Univ Tokyo Tokyo 1138654 Japan RIKEN Ctr Adv Intelligence Project Tokyo 1030027 Japan Johns Hopkins Univ Appl Phys Lab Laurel MD 11100 USA European Space Res Inst & Lab European Space Agcy I-00044 Frascati Italy

The five papers in this special section focus on computer vision-based approaches for Earth observation. These papers followed a series of events promoting works at the interface between computer vision and remote sensing: the special sessions organized at the Living Planet Symposium1 and the computer vision and pattern recognition (cvpr) conference (the Earthvision2 and computer vision for Global Challenges3 workshops). These sessions aimed at fostering collaboration between the computer vision and earth observation communities to boost automated interpretation of remotely sensed data. They also aimed at raising awareness inside the computer vision community for this highly challenging and quickly evolving field of research with a big impact on human society, economy, industry, and the planet. Submissions were invited from all areas of computer vision and image analysis relevant for, or applied to environmental remote sensing and were not limited to the papers presented at the events above. The papers retained in this special issue reflect the high variety of automatic image analysis in remote sensing.

关键词： Special issues and sections Remote sensing computer vision Satellites Hyperspectral images Synthetic aperture radar

来源：评论

学校读者我要写书评

暂无评论

Bilinear Parameterization For Differentiable Rank-Regularization

Bilinear Parameterization For Differentiable Rank-Regulariza...

引用

ieee/CVF conference on computer vision and pattern recognition (cvpr)

作者： Ornhag, Marcus Valtonen Olsson, Carl Heyden, Anders Lund Univ Ctr Math Sci Lund Sweden Chalmers Univ Technol Dept Elect Engn Gothenburg Sweden

ISBN: (纸本)9781728193601

Low rank approximation is a commonly occurring problem in many computer vision and machine learning applications. There are two common ways of optimizing the resulting models. Either the set of matrices with a given rank can be explicitly parametrized using a bilinear factorization, or low rank can be implicitly enforced using regularization terms penalizing non-zero singular values. While the former approach results in differentiable problems that can be efficiently optimized using local quadratic approximation, the latter is typically not differentiable (sometimes even discontinuous) and requires first order subgradient or splitting methods. It is well known that gradient based methods exhibit slow convergence for ill-conditioned problems. In this paper we show how many non-differentiable regularization methods can be reformulated into smooth objectives using bilinear parameterization. This allows us to use standard second order methods, such as Levenberg-Marquardt (LM) and Variable Projection (VarPro), to achieve accurate solutions for ill-conditioned cases. We show on several real and synthetic experiments that our second order formulation converges to substantially more accurate solutions than competing state-of-the-art methods.(1)

关键词： computer vision

来源：评论

学校读者我要写书评

暂无评论

DNDNet: Reconfiguring CNN for Adversarial Robustness

DNDNet: Reconfiguring CNN for Adversarial Robustness

引用

ieee/CVF conference on computer vision and pattern recognition (cvpr)

作者： Goel, Akhil Agarwal, Akshay Vatsa, Mayank Singh, Richa Ratha, Nalini K. IIIT Delhi Delhi India IIT Jodhpur Jodhpur Rajasthan India IBM TJ Watson Res Ctr Yorktown Hts NY USA

ISBN: (数字)9781728193601

ISBN: (纸本)9781728193601

Several successful adversarial attacks have demonstrated the vulnerabilities of deep learning algorithms. These attacks are detrimental in building deep learning based dependable AI applications. Therefore, it is imperative to build a defense mechanism to protect the integrity of deep learning models. In this paper, we present a novel "defense layer" in a network which aims to block the generation of adversarial noise and prevents an adversarial attack in black-box and gray-box settings. The parameter-free defense layer, when applied to any convolutional network, helps in achieving protection against attacks such as FGSM, L2, Elastic-Net, and DeepFool. Experiments are performed with different CNN architectures, including VGG, ResNet, and DenseNet, on three databases, namely, MNIST, CIFAR-10, and PaSC. The results showcase the efficacy of the proposed defense layer without adding any computational overhead. For example, on the CIFAR-10 database, while the attack can reduce the accuracy of the ResNet-50 model to as low as 6.3%, the proposed "defense layer" retains the original accuracy of 81.32%.

关键词： Mathematical model Perturbation methods Machine learning computer architecture Robustness Computational modeling Databases

来源：评论

学校读者我要写书评

暂无评论

Discriminant Distribution-Agnostic Loss for Facial Expression recognition in the Wild

Discriminant Distribution-Agnostic Loss for Facial Expressio...

引用

ieee/CVF conference on computer vision and pattern recognition (cvpr)

作者： Farzaneh, Amir Hossein Qi, Xiaojun Utah State Univ Dept Comp Sci Logan UT 84322 USA

ISBN: (纸本)9781728193601

Facial Expression recognition (FER) has demonstrated remarkable progress due to the advancement of deep Convolutional Neural Networks (CNNs). FER's goal as a visual recognition problem is to learn a mapping from the facial embedding space to a set of fixed expression categories using a supervised learning algorithm. Softmax loss as the de facto standard in practice fails to learn discriminative features for efficient learning. Center loss and its variants as promising solutions increase deep feature discriminability in the embedding space and enable efficient learning. They fundamentally aim to maximize intra-class similarity and inter-class separation in the embedding space. However, center loss and its variants ignore the underlying extreme class imbalance in challenging wild FER datasets. As a result, they lead to a separation bias toward majority classes and leave minority classes overlapped in the embedding space. In this paper, we propose a novel Discriminant Distribution-Agnostic loss (DDA loss) to optimize the embedding space for extreme class imbalance scenarios. Specifically, DDA loss enforces inter-class separation of deep features for both majority and minority classes. Any CNN model can be trained with the DDA loss to yield well separated deep feature clusters in the embedding space. We conduct experiments on two popular large-scale wild FER datasets (RAF-DB and AffectNet) to show the discriminative power of the proposed loss function.

关键词： Face recognition

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：