检索结果-内蒙古大学图书馆

Semi-Supervised Class-Agnostic Motion Prediction with Pseudo lab.l Regeneration and BEVMix

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Wang, Kewei Wu, Yizheng Pan, Zhiyu Li, Xingyi Xian, Ke Wang, Zhe Cao, Zhiguo Lin, Guosheng S-Lab Nanyang Technological University Singapore Key Laboratory of Image Processing and Intelligent Control Ministry of Education School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan430074 China SenseTime Research Hong Kong

Class-agnostic motion prediction methods aim to comprehend motion within open-world scenarios, holding significance for autonomous driving systems. However, training a high-performance model in a fully-supervised manner always requires substantial amounts of manually annotated data, which can be both expensive and time-consuming to obtain. To address this challenge, our study explores the potential of semi-supervised learning (SSL) for class-agnostic motion prediction. Our SSL framework adopts a consistency-based self-training paradigm, enabling the model to learn from unlab.led data by generating pseudo lab.ls through test-time inference. To improve the quality of pseudo lab.ls, we propose a novel motion selection and re-generation module. This module effectively selects reliable pseudo lab.ls and regenerates unreliable ones. Furthermore, we propose two data augmentation strategies: temporal sampling and BEVMix. These strategies facilitate consistency regularization in SSL. Experiments conducted on nuScenes demonstrate that our SSL method can surpass the self-supervised approach by a large margin by utilizing only a tiny fraction of lab.led data. Furthermore, our method exhibits comparable performance to weakly and some fully supervised methods. These results highlight the ability of our method to strike a favorable balance between annotation costs and performance. Code will be availab.e at https://***/kwwcv/SSMP. Copyright © 2023, The Authors. All rights reserved.

关键词： Forecasting

Adversarial refinement network for human motion prediction

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Chao, Xianjin Bin, Yanrui Chu, Wenqing Cao, Xuan Ge, Yanhao Wang, Chengjie Li, Jilin Huang, Feiyue Leung, Howard City University of Hong Kong Hong Kong Hong Kong Key Laboratory of Image Processing and Intelligent Control School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan China Tencent Youtu Lab Shanghai China

Human motion prediction aims to predict future 3D skeletal sequences by giving a limited human motion as inputs. Two popular methods, recurrent neural networks and feed-forward deep networks, are able to predict rough motion trend, but motion details such as limb movement may be lost. To predict more accurate future human motion, we propose an Adversarial Refinement Network (ARNet) following a simple yet effective coarse-to-fine mechanism with novel adversarial error augmentation. Specifically, we take both the historical motion sequences and coarse prediction as input of our cascaded refinement network to predict refined human motion and strengthen the refinement network with adversarial error augmentation. During training, we deliberately introduce the error distribution by learning through the adversarial mechanism among different subjects. In testing, our cascaded refinement network alleviates the prediction error from the coarse predictor resulting in a finer prediction robustly. This adversarial error augmentation provides rich error cases as input to our refinement network, leading to better generalization performance on the testing dataset. We conduct extensive experiments on three standard benchmark datasets and show that our proposed ARNet outperforms other state-of-the-art methods, especially on challenging aperiodic actions in both short-term and long-term predictions. © 2020, CC-BY.

关键词： Errors

Determination of pitch range based on onset and offset analysis in modulation frequency domain

学校读者我要写书评

暂无评论

2010 5th International Symposium on Telecommunications, IST ...

2010 5th International Symposium on Telecommunications, IST 2010 2010年 604-608页

作者： Mahmoodzadeh, A. Abutalebi, H.R. Soltanian-Zadeh, H. Sheikhzadeh, H. Speech Proc. Research Lab. ECE Dept. Yazd University Yazd Iran Control and Intelligent Processing Center of Excellence University of Tehran Tehran Iran EE Dept. Amirkabir University of Technology Tehran Iran Idiap Research Institute Martigny Switzerland Image Analysis Lab. Henry Ford Health System Detroit United States

ISBN: (纸本)9781424481835

Auditory scene in a natural environment contains multiple sources. Auditory scene analysis (ASA) is the process in which the auditory system segregates a scene into streams corresponding to different sources. The determination of range of pitch frequency is necessary for segmentation. We propose a system to determine the range of pitch frequency by analyzing onsets and offsets in modulation frequency domain. In the proposed system, first the modulation spectrum of speech is calculated and then, in each subband onsets and offsets will be detected. Thereafter, the segments are generated by matching corresponding onset and offset front. Finally, by choosing the desired segments, the rage of pitch frequency is determined. Systematic evaluation shows that the range of pitch frequency is estimated with good accuracy. © 2010 IEEE.

关键词： Frequency modulation

EDGE DETECTION USING SUPPORT VALUE TRANSFROM

学校读者我要写书评

暂无评论

EDGE DETECTION USING SUPPORT VALUE TRANSFROM

2007 International Conference on Machine Learning and Cybernetics(IEEE第六届机器学习与控制论国际会议)

作者： SHENG ZHENG JINWEN TIAN JIAN LIU Electronics and Information Engineering Department Huazhong University of Science and Technology W State Education Commission Key Lab for Image Processing and Intelligent Control HUST Wuhan 430074 Electronics and Information Engineering Department Huazhong University of Science and Technology W

An innovative edge detection algorithm using the support value transform is presented in this *** on the support value transform, the multi scale support value images are extracted from the *** resolution of an image is directly related to the proper scale for edge detection, and the second level support value image is used to locate the edge positions by their zero *** experiments are carried out and sharp image edges are obtained from a variety of sample *** with many other existing methods, including LoG and Canny detectors, the proposed algorithm is superior to the LOG and Roberts approach.

关键词： Edge detection mapped least squares support vector machine (mapped LS-SVM) support value transform (SVT) zero crossings

Discovering true association between multimodal data sets using structured and sparse canonical correlation analysis: A simulation study

学校读者我要写书评

暂无评论

Discovering true association between multimodal data sets us...

IEEE International Symposium on Biomedical Imaging

作者： Ali-Reza Mohammadi-Nejad Gholam-Ali Hossein-Zadeh Hamid Soltanian-Zadeh Control and Intelligent Processing Center of Excellence (CIPCE) University of Tehran Tehran Iran School of Cognitive Sciences Institute for Research in Fundamental Sciences (IPM) Tehran Iran Henry Ford Health System Medical Image Analysis Lab. Detroit Michigan USA

ISBN: (纸本)9781479923519

Recently, multi-modal data fusion emerged as a comprehensive approach in the field of neuroimaging. As a multivariate approach, canonical correlation analysis (CCA) is frequently used for data fusion. However, the current CCA-based fusion approaches face many problems in the sense of high-dimensionality, multi-colinearity, unimodal feature selection, asymmetric and reshaping of the imaging data into vectors. In this paper, we propose a structured and sparse CCA (ssCCA) technique to overcome the above problems. To investigate the performance of the proposed approach, we compared three multivariate and multi-modal fusion techniques: standard CCA, regularized sparse CCA, and ssCCA, with respect to their ability to detect multi-modal data associations. The results demonstrate that the ssCCA approach has superior recovery performance in the extraction of the true correlated variables relative to the other two fusion approaches.

关键词： Correlation Imaging Laplace equations Data integration Data models Standards Tuning

Data-driven Sparse Bayesian Learning Discovery of Hysteresis Effects in Electrospinning Processes

学校读者我要写书评

暂无评论

Data-driven Sparse Bayesian Learning Discovery of Hysteresis...

第34届中国控制与决策会议

作者： Hongwei Sun Yasen Wang Yaozhong Zheng Xiang Huang Hai-Tao Zhang School of Artificial Intelligence and Automation and Key Laboratory of Image Processing and Intelligent ControlSchool of AutomationHuazhong University of Science and Technology State Key Lab of Digital Manufacturing Equipment and Technology Huazhong University of Science and Technology School of Mechanical Science and Engineering Huazhong University of Science and Technology

Electrospinning is an efficient and feasible method to fabricate hairline fibers from polymers or ***,the hysteresis nonlinearity among pump flow and jet diameter during electrospinning process strongly hinders the improvement of its *** alleviate this effect,it is an urgent yet challenging mission to discover hysteresis effects in near-field Electrospinning *** this work,a nonlinear autoregressive exogenous model is developed for described electrospinning ***,a pure data-driven sparse B ayesian learning(SBL) method is applied to distill the pump flow-jet diameter hysteresis ***,extensive experiments are conducted to verify the discovered effects by the SBL identification method.

关键词： Electrospinning hysteresis nonlinear autoregressive exogenous(NARX) sparse Bayesian learning(SBL)

Automatic needle segmentation using 3D Gray-level Hough Transform in 3D ultrasound images

学校读者我要写书评

暂无评论

Journal of Computational Information Systems 2008年第5期4卷 2021-2026页

作者： Qiu, Wu Ding, Mingyue Yuchi, Ming Image Processing and Intelligence Control Key Lab. of Education Ministry of China Institute of Pattern Recognition and Artificial Intelligence Huazhong University of Science and Technology Wuhan 430074 China Department of Biomedical Engineering School of Life Science and Technology Huazhong University of Science and Technology Wuhan 430074 China

This paper describes a novel 3D needle segmentation algorithm for 3DUS data. The algorithm includes the 3D Gray-level Hough Transform (3DGHT), which is based on the representation (ψ, θ, ρ, α) of straight lines in 3D space and Local Dimensionality Estimator (GST). Different from the traditional Hough transformation the process of binarization is not needed, because the proposed algorithm adopts the original gray level image directly. We tested it with the synthetic data and water phantom. The results show that our algorithm works well in 3D US images with angular deviation less than 1° and position deviation less than 1 mm. © 2008 Binary Information Press.

关键词： Hough transforms

Temporally-Adaptive Models for Efficient Video Understanding

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Huang, Ziyuan Zhang, Shiwei Pan, Liang Qing, Zhiwu Zhang, Yingya Liu, Ziwei Ang, Marcelo H. Advanced Robotics Centre National University of Singapore Singapore DAMO Academy Alibaba Group China S-Lab National Technological University Singapore Key Laboratory of Image Processing and Intelligent Control School of Artificial Intelligence and Automation Huazhong University of Science China

Spatial convolutions are extensively used in numerous deep video models. It fundamentally assumes spatio-temporal invariance, i.e., using shared weights for every location in different frames. This work presents Temporally-Adaptive Convolutions (TAdaConv) for video understanding, which shows that adaptive weight calibration along the temporal dimension is an efficient way to facilitate modeling complex temporal dynamics in videos. Specifically, TAdaConv empowers spatial convolutions with temporal modeling abilities by calibrating the convolution weights for each frame according to its local and global temporal context. Compared to existing operations for temporal modeling, TAdaConv is more efficient as it operates over the convolution kernels instead of the features, whose dimension is an order of magnitude smaller than the spatial resolutions. Further, kernel calibration brings an increased model capacity. Based on this readily plug-in operation TAdaConv as well as its extension, i.e., TAdaConvV2, we construct TAdaBlocks to empower ConvNeXt and Vision Transformer to have strong temporal modeling capabilities. Empirical results show TAdaConvNeXtV2 and TAdaFormer perform competitively against state-of-the-art convolutional and Transformer-based models in various video understanding benchmarks. Our codes and models are released at: https://***/alibaba-mmai-research/TAdaConv. Copyright © 2023, The Authors. All rights reserved.

关键词： Convolution

Universal Adversarial Perturbations for CNN Classifiers in EEG-Based BCIs

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Liu, Zihan Meng, Lubin Zhang, Xiao Fang, Weili Wu, Dongrui Key Laboratory of the Ministry of Education for Image Processing and Intelligent Control School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan430074 China School of Design and Environment National University of Singapore 117566 Singapore Zhejiang Lab Hangzhou311121 China

Multiple convolutional neural network (CNN) classifiers have been proposed for electroencephalogram (EEG) based brain-computer interfaces (BCIs). However, CNN models have been found vulnerable to universal adversarial perturbations (UAPs), which are small and example-independent, yet powerful enough to degrade the performance of a CNN model, when added to a benign example. This paper proposes a novel total loss minimization (TLM) approach to generate UAPs for EEG-based BCIs. Experimental results demonstrated the effectiveness of TLM on three popular CNN classifiers for both target and non-target attacks. We also verified the transferability of UAPs in EEG-based BCI systems. To our knowledge, this is the first study on UAPs of CNN classifiers in EEG-based BCIs. UAPs are easy to construct, and can attack BCIs in real-time, exposing a potentially critical security concern of BCIs. Copyright © 2019, The Authors. All rights reserved.

关键词： Electroencephalography