Filters








3,963 Hits in 4.9 sec

Probabilistic Video Generation using Holistic Attribute Control [article]

Jiawei He, Andreas Lehrmann, Joseph Marino, Greg Mori, Leonid Sigal
2018 arXiv   pre-print
We improve the video generation consistency through temporally-conditional sampling and quality by structuring the latent space with attribute controls; ensuring that attributes can be both inferred and  ...  Based on this intuition, we propose a generative framework for video generation and future prediction.  ...  Fig. 1 : Video Generation using Attribute Control. Our framework uses a semi-supervised latent space containing a fixed number of control signals to steer the generation.  ... 
arXiv:1803.08085v1 fatcat:nzylgu2tyjgr7frm5ismzoze5i

Face Recognition by Computers and Humans

Rama Chellappa, Pawan Sinha, P. Jonathon Phillips
2010 Computer  
Algorithm Short description Experimental description Probabilistic recognition of human faces from video x Simultaneous tracking and recognition using a dynamic state space model and sequential  ...  local attributes.  ... 
doi:10.1109/mc.2010.37 fatcat:lpttfczwk5g2boj2f7chtx6r6i

Facial Landmark Detection: A Literature Survey

Yue Wu, Qiang Ji
2018 International Journal of Computer Vision  
We also compare their performances on both controlled and in the wild benchmark datasets, under varying facial expressions, head poses, and occlusion.  ...  The holistic methods explicitly build models to represent the global facial appearance and shape information.  ...  Databases under "controlled" conditions Databases under "controlled" conditions refer to databases with video/images collected indoor with certain restrictions (e.g. pre-defined expressions, head poses  ... 
doi:10.1007/s11263-018-1097-z fatcat:ykqg6lr3j5bbrmrmli2dlrxupi

Exemplar Hidden Markov Models for classification of facial expressions in videos

Karan Sikka, Abhinav Dhall, Marian Bartlett
2015 2015 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)  
A probabilistic kernel is then used to compute a kernel matrix, to be used along with an SVM classifier.  ...  to the challenges of estimating generative models.  ...  As a result, generative HMMs generally have a lower performance compared to Disc S-T approaches [9, 24] , and are seldom used despite their modeling capabilities.  ... 
doi:10.1109/cvprw.2015.7301350 dblp:conf/cvpr/SikkaDB15 fatcat:bwaxxpbvqvaxlp3x7ekzguipee

Video Captioning with Transferred Semantic Attributes [article]

Yingwei Pan, Ting Yao, Houqiang Li, Tao Mei
2016 arXiv   pre-print
Automatically generating natural language descriptions of videos plays a fundamental challenge for computer vision community.  ...  To boost video captioning, we propose a novel transfer unit to model the mutually correlated attributes learnt from images and videos.  ...  To verify our claim, we have presented video MIL framework to holistically explore semantic information in a video and a transfer unit to contextually control the impacts of attributes learnt from images  ... 
arXiv:1611.07675v1 fatcat:eb6u7yq6fnc4vllywsrlqhf7hy

Video Captioning with Transferred Semantic Attributes

Yingwei Pan, Ting Yao, Houqiang Li, Tao Mei
2017 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)  
Automatically generating natural language descriptions of videos plays a fundamental challenge for computer vision community.  ...  To boost video captioning, we propose a novel transfer unit to model the mutually correlated attributes learnt from images and videos.  ...  To verify our claim, we have presented video MIL framework to holistically explore semantic information in a video and a transfer unit to contextually control the impacts of attributes learnt from images  ... 
doi:10.1109/cvpr.2017.111 dblp:conf/cvpr/PanYLM17 fatcat:mlmll73movgs5hacijd4omiyve

A Survey of Deep Facial Attribute Analysis [article]

Xin Zheng, Yanqing Guo, Huaibo Huang, Yi Li, Ran He
2019 arXiv   pre-print
Second, the datasets and performance metrics commonly used in facial attribute analysis are presented.  ...  First, we summarize a general pipeline that deep facial attribute analysis follows, which comprises two stages: data preprocessing and model construction.  ...  [72] propose a probabilistic confidence criterion to address this inconsistency issue.  ... 
arXiv:1812.10265v3 fatcat:tezgo2angvfefbttuoodnss6t4

Crowd Analysis and Its Applications [chapter]

Nilam Nur Amir Sjarif, Siti Mariyam Shamsuddin, Siti Zaiton Mohd Hashim, Siti Sophiayati Yuhaniz
2011 Communications in Computer and Information Science  
In this paper, we give the general framework and taxonomy of pattern in detecting abnormal behavior in a crowd scene.  ...  Meanwhile, the common process for analysis in video sequence of crowd information extraction consists of Pre-Processing, Object Tracking, and Event/Behavior Recognition.  ...  CCTV is used to observe parts of a process from control environment which is required in every intelligent crowded scene.  ... 
doi:10.1007/978-3-642-22170-5_59 fatcat:s33xn5ez5bev7gnl4jhied73sm

A Survey of Deep Facial Attribute Analysis

Xin Zheng, Yanqing Guo, Huaibo Huang, Yi Li, Ran He
2020 International Journal of Computer Vision  
Second, the datasets and performance metrics commonly used in facial attribute analysis are presented.  ...  First, we summarize a general pipeline that deep facial attribute analysis follows, which comprises two stages: data preprocessing and model construction.  ...  However, ResGAN generates residual images for locating attribute-relevant regions under the sparsity constraint. Such a constraint relies heavily on control parameters but not attributes themselves.  ... 
doi:10.1007/s11263-020-01308-z fatcat:xmlukvd5qbenzkzjacefhcnope

Spatio-Temporal Action Localization in a Weakly Supervised Setting [article]

Kurt Degiorgio, Fabio Cuzzolin
2019 arXiv   pre-print
Subsequently, a convolutional neural network is used to extract RGB features from the resulting video segments.  ...  However, the data requirements needed to achieve adequate generalization in this setting is prohibitive.  ...  The novelty element in our work is the fact that we use a probabilistic MIL formulation to generalize over new features, in conjunction with a video splitting technique applied at training time, that allows  ... 
arXiv:1905.02171v1 fatcat:w62ivqe7pnh4hex2po632cu7ja

Crowd Counting via Weighted VLAD on Dense Attribute Feature Maps [article]

Biyun Sheng, Chunhua Shen, Guosheng Lin, Jun Li, Wankou Yang, Changyin Sun
2016 arXiv   pre-print
Conventional holistic features used in crowd counting often fail to capture semantic attributes and spatial cues of the image.  ...  First, with the help of convolutional neural network (CNN), the original pixel space is mapped onto a dense attribute feature map, where each dimension of the pixel-wise feature indicates the probabilistic  ...  In recent years, attribute-based representations that describe a target by multiple attribute classes have been widely used to represent objects [24] , faces [25] and actions [26] .  ... 
arXiv:1604.08660v1 fatcat:o4odtwqpvrgpvbzbzrnqmaqo7y

An Evaluation of Video-to-Video Face Verification

Norman Poh, Chi Ho Chan, Josef Kittler, Sébastien Marcel, Christopher Mc Cool, Enrique Argones Rua, José Luis Alba Castro, Mauricio Villegas, Roberto Paredes, Vitomir Struc, Nikola Pavesic, Albert Ali Salah (+2 others)
2010 IEEE Transactions on Information Forensics and Security  
This paper presents an evaluation of person identity verification using facial video data, organized in conjunction with the International Conference on Biometrics (ICB 2009).  ...  However, due to the widespread use of web-cams and mobile devices embedded with a camera, it is now possible to realize facial video recognition, rather than resorting to just still images.  ...  The problem of this generative model is that it has a limited discriminatory capacity. The work in [58] extends [36] to allow on-line learning of probabilistic appearance manifolds.  ... 
doi:10.1109/tifs.2010.2077627 fatcat:vfbkbo7gavh7lboxlx3pboic6q

Automatic multi-view face recognition via 3D model based pose regularization

Koichiro Niinuma, Hu Han, Anil K. Jain
2013 2013 IEEE Sixth International Conference on Biometrics: Theory, Applications and Systems (BTAS)  
We first build a 3D model from each frontal target face image, which is used to generate synthetic target face images.  ...  The pose of a query face image is also estimated using a multi-view face detector so that the synthetic target face images can be generated to resemble the pose variation of a query face image.  ...  To replicate the scenarios of face recognition from images or videos captured using mobile devices, we have collected a Mobile dataset consisting of 112 subjects using iPhone 4S.  ... 
doi:10.1109/btas.2013.6712735 dblp:conf/btas/NiinumaHJ13 fatcat:m2ixkw747jedjiui4xk5kx2yle

Transformative art: art as means for long-term neurocognitive change

Son Preminger
2012 Frontiers in Human Neuroscience  
Indeed, artworks require the experiencer to "complete the experience" using imagination and other internally generated cognitive processes to fill in the gap to create a holistic experience.  ...  It has been suggested that probabilistic inference may be a general learning mechanism underlying these wide-range improvements (Green et al., 2010) .  ... 
doi:10.3389/fnhum.2012.00096 pmid:22536178 pmcid:PMC3334843 fatcat:42jruabq7vgwjdsouyfq6h7imy

Vision based Traffic Police Hand Signal Recognition in Surveillance Video - A Survey

R. Sathya, M. Kalaiselvi Geetha
2013 International Journal of Computer Applications  
General overview of an traffic control gestures and its various applications where discussed in this paper.  ...  Most of the recognition system uses the benchmark datasets like KTH, Weizmann. some other datasets were used by the action recognition system.  ...  Models for generally into the class of graphical models, which are best described as probabilistic grammars.  ... 
doi:10.5120/14037-2192 fatcat:dtns3iu3fje77dnrgsn2346qoq
« Previous Showing results 1 — 15 out of 3,963 results