Description: Most motion-based tracking algorithms assume that objects undergo rigid motion, which is most likely disobeyed in real world. In this paper, we present a novel motion-based tracking framework which makes no such assumptions. Object is represented by a set of local invariant features, whose motions are observed by a feature correspon-
dence process. A generative model is proposed to depict
the relationship between local feature motions and object
global motion, whose parameters are learned efciently by
an on-line EM algorithm. And the object global motion is estimated in term of maximum likelihood of observations.Then an updating mechanism is employed to adapt object representation. Experiments show that our framework is
exible and robust in dealing with appearance changes,background clutter, illumination changes and occlusion Platform: |
Size: 424705 |
Author:chenjieke |
Hits:
Description: 汽车高斯曲面拟合
---
2程序，以适应到表面二维高斯：
子= A *的进出口（ -（（西为X0）^2/2/sigmax^2 +（艺Y0的）^2/2/sigmay^ 2）。。）+ b的
这些例程是自动在某种意义上说，他们并不需要出发对模型参数的猜测规范。
autoGaussianSurfML（十一，彝，子）适合通过对模型参数的最大似然（最小二乘）。它首先计算了该模型在许多可能的参数值，然后选择最佳质量设置和细化与lsqcurvefit它。
autoGaussianSurfGS（十一，彝，紫）的估计，通过指定数据的贝叶斯生成模型，然后采取通过从模型吉布斯抽样样本后ofthis模型参数。这种-Auto Gaussian Surface fit
---
2 routines to fit a 2D Gaussian to a surface:
zi = a*exp(-((xi-x0).^2/2/sigmax^2+ (yi-y0).^2/2/sigmay^2))+ b
The routines are automatic in the sense that they do not require the specification of starting guesses for the model parameters.
autoGaussianSurfML(xi,yi,zi) fits the model parameters through maximum likelihood(least-squares). It first evaluates the quality of the model at many possible values of the parameters then chooses the best set and refines it with lsqcurvefit.
autoGaussianSurfGS(xi,yi,zi) estimates the model parameters by specifying a Bayesian generative model for the data, then taking samples from the posterior ofthis model through Gibbs sampling. This method is insensitive to local minimain posterior and gives meaningful error bars (Bayesian confidence intervals) Platform: |
Size: 7168 |
Author:zzskzcau |
Hits:
Description: Given two to four synchronized video streams taken at eye level and from different angles, we show
that we can effectively combine a generative model with dynamic programming to accurately follow
up to six individuals across thousands of frames in spite of significant occlusions and lighting changes.
In addition, we also derive metrically accurate trajectories for each one of them.
Platform: |
Size: 1384448 |
Author:pierounix85 |
Hits:
Description: 一本非常好的关于ITU解说的书籍，通过它可以了解ITU信道的生成原理。-A very good about ITU explain books, by which it can understand the ITU channel generative principle. Platform: |
Size: 397312 |
Author:Wulong |
Hits:
Description: The 3-D Morphable Model was
introduced as a generative model to p redictthe appearances o f
an individual while using a statistical prior on shape and
texture allowin g its parameters to be estimated from single
image. Based on these new unde rstandings , face recognition
algorithms have been developed to address the joint challenges of pose and lighting. Platform: |
Size: 1209344 |
Author:bobobobo |
Hits:
Description: 这是一篇关于多个物体类的协同分割应用到视频分割的论文。-Multi-Class Video Co-Segmentation with a Generative Multi-Video Model ；2013 IEEE Conference on Computer Vision and Pattern Recognition Platform: |
Size: 1651712 |
Author:jackson |
Hits:
Description: LDA是一种文档主题生成模型，也称为一个三层贝叶斯概率模型，包含词、主题和文档三层结构。文档到主题服从Dirichlet分布，主题到词服从多项式分布。
LDA是一种非监督机器学习技术，可以用来识别大规模文档集（document collection）或语料库（corpus）中潜藏的主题信息。它采用了词袋（bag of words）的方法，这种方法将每一篇文档视为一个词频向量，从而将文本信息转化为了易于建模的数字信息。但是词袋方法没有考虑词与词之间的顺序，这简化了问题的复杂性，同时也为模型的改进提供了契机。每一篇文档代表了一些主题所构成的一个概率分布，而每一个主题又代表了很多单词所构成的一个概率分布。
对于语料库中的每篇文档，LDA定义了如下生成过程（generative process）：
1. 对每一篇文档，从主题分布中抽取一个主题；
2. 从上述被抽到的主题所对应的单词分布中抽取一个单词；
3. 重复上述过程直至遍历文档中的每一个单词。-LDA is a document theme generation model, also known as a three-tier Bayesian probability model that contains the words, topics and document three-tier structure. Dirichlet distribution of the document to the theme of obedience, the theme to the word obey polynomial distribution.
LDA is an unsupervised machine learning techniques can be used to identify large-scale document set (document collection) or corpus (corpus) of the underlying themes of information. It uses the word bag (bag of words) of the method, which each one document as a word frequency vector, thus the text information into digital information for ease of modeling. However, the method does not consider the order of the words Bag between words, which simplifies the complexity of the problem, but also for the improvement of the model provided an opportunity. Each document represents a probability distribution of some of the topics posed, and each topic and they represent many words constituted a probability distribut Platform: |
Size: 30720 |
Author:yangling |
Hits:
Description: Kalman Filters are one of the most influential models of time-varying phenomena.They admit an intuitive probabilistic interpretation, have a simple functional form,and enjoy widespread adoption in a variety of disciplines. Motivated by recentvariational methods for learning deep generative models, we introduce a unifiedalgorithm to efficiently learn a broad spectrum of Kalman filters. Of particularinterest is the use of temporal generative models for counterfactual inference. Weinvestigate the efficacy of such models for counterfactual inference, and to thatend we introduce the “Healing MNIST” dataset where long-term structure, noiseand actions are applied to sequences of digits. We show the efficacy of our methodfor modeling this dataset. We further show how our model can be used for coun-terfactual inference for patients, based on electronic health record data of 8,000patients over 4.5 years. Platform: |
Size: 1693696 |
Author:Lea |
Hits:
Description: Matlab implementation of the EM and MCMC algorithm for SVMs as introduced in the paper Data augmentation for support vector machines http://ba.stat.cmu.edu/journal/2011/vol06/issue01/polson.pdf-This is a Matlab implementation of the fancy idea by Polson & Scott that reformulates the traditional binary linear SVM problem into a MAP (Maximum a Posteriori) estimation in a probabilistic generative model, and by use of the technique of data augmentation, makes it possible to do very easy and fast Gibbs sampling for the solution. Platform: |
Size: 7168 |
Author:saisai |
Hits:
Description: 本文是关于机器学习的全面介绍，主要涵盖内容有：监督学习，生成学习算法，支持向量机，学习理论，EM算法，感知器和大幅度分类器，正则化和模型选择等-this paper is about machine learning,including:Supervised learning、Generative Learning algorithms、Support Vector Machines、Learning Theory、The EM algorithm、The perceptron and large margin classifiers、Regularization and model selection Platform: |
Size: 3403776 |
Author:高乐莲 |
Hits:
Description: The number of states in GMM as the generative model of the frames is obtained using
k-means algorithm. This also helps to initialize the mean vector and the covariance
matrix of the individual state of the GMM. The training LPC frames collected from
three speech segments are subjected to PCA for dimensionality reduction and are
subjected to k-means algorithm. The total number of frames is equal to the total
number of vectors that are subjected to k-means clustering. Platform: |
Size: 728064 |
Author:Khan17
|
Hits:
Description: 代码适用于追踪物体，可以很有效的图中进行精确的实时追踪(This code impelemts a method for tracking an object in a sequence of images given its location in the first frame. In this approach, a combination of generative and discriminative methods is used to model the object appearance.) Platform: |
Size: 4141056 |
Author:懒样
|
Hits:
Description: 此程序为对抗生成网络，生成图像。
生成对抗网络是一种生成模型（Generative Model），其背后基本思想是从训练库里获取很多训练样本，从而学习这些训练案例生成的概率分布。
而实现的方法，是让两个网络相互竞争，‘玩一个游戏’。其中一个叫做生成器网络（ Generator Network），它不断捕捉训练库里真实图片的概率分布，将输入的随机噪声（Random Noise）转变成新的样本（也就是假数据）。另一个叫做判别器网络（Discriminator Network），它可以同时观察真实和假造的数据，判断这个数据到底是不是真的。”(This program is Generative Adversarial Net,and generate images.
The generation of confrontation network is a generative model (Generative Model). The basic idea behind it is to get many training samples from the training library, so as to learn the probability distribution of these training cases.
The realization of the method is to let the two networks compete with each other, 'play a game'. One is called Generator Network, which constantly captures the probability distribution of real pictures in training libraries, and transforms the input random noise (Random Noise) into new samples (that is, false data). Another method is called network (Discriminator Network), it can observe the true and false data, determine the data in the end is not really.") Platform: |
Size: 169984 |
Author:王小西 |
Hits:
Description: 生成式对抗网络（GAN, Generative Adversarial Networks ）是一种深度学习模型，是近年来复杂分布上无监督学习最具前景的方法之一。模型通过框架中（至少）两个模块：生成模型（Generative Model）和判别模型（Discriminative Model）的互相博弈学习产生相当好的输出。(Emergent against network (GAN, Generative Adversarial Networks) is a kind of deep learning model, is a complex distribution in recent years on one of the most promising method for unsupervised learning.
The frame of the Model by (at least) two modules: generation Model (Generative Model) and the discriminant Model (Discriminative Model) of the game to learn each other produce fairly good output.) Platform: |
Size: 6144 |
Author:yaofang123 |
Hits:
Description: 本书由Keras之父、现任Google A工智能研究员的弗朗索瓦?肖莱（Frangois Chollet）执笔，详尽介 绍了用Python和Keras进行深度学习的探索实践，涉及计算机视觉、自然语言处理、生成式模型等应用。 书中包含30多个代码示例，步骤讲解详细透彻。由于本书立足于人工智能的可达性和大众化，读者无须 具备机器学习相关背景知识即可展开阅读。在学习完本书后，读者将具备搭建自己的深度学习环境、建立 图像识别模型、生成图像和文字等能力(This book is written by Frangois Chollet, the father of keras and the current researcher of Google a intelligence. It introduces in detail the exploration and practice of deep learning with Python and keras, involving computer vision, natural language processing, generative model and other applications. The book contains more than 30 code examples, and the steps are detailed and thorough. Because this book is based on the accessibility and popularization of artificial intelligence, readers can read it without having the background knowledge of machine learning. After learning this book, readers will have the ability to build their own deep learning environment, establish image recognition model, and generate images and characters) Platform: |
Size: 2529280 |
Author:jliop |
Hits:
Description: Digital watermark embeds informa-
tion bits into digital cover such as images and
videos to prove the creator’s ownership of his
work. In this paper, we propose a robust image
watermark algorithm based on a generative
adversarial network. This model includes two
modules, generator and adversary. Generator
is mainly used to generate images embedded
with watermark, and decode the image dam-
aged by noise to obtain the watermark. Adver-
sary is used to discriminate whether the image
is embedded with watermark and damage the
image by noise. Based on the model Hidden
(hiding data with deep networks), we add a
high-pass filter in front of the discriminator,
making the watermark tend to be embedded in
the mid-frequency region of the image. Since
the human visual system pays more attention
to the central area of the image, we give a
higher weight to the image center region, and
a lower weight to the edge region when calcu-
lating the loss between cover and embedded
image. The watermarked image obtained by
this scheme has a better visual performance.
Experimental results show that the proposed
architecture is more robust against noise
interference compared with the state-of-art
schemes. Platform: |
Size: 704508 |
Author:bamzi334 |
Hits:
Search - generative model