Alternating Feature Spaces in Relevance Feedback

Image retrieval using relevance feedback can be treated as a two-class learning and classification process. The user-labelled relevant and irrelevant images are regarded as positive and negative training samples, based on which a classifier is trained dynamically. Then the classifier in turn classifies all images in the database. In practice, the number of training samples is very small because the users are often impatient. On the other hand, the positive samples usually are not representative since they are the nearest ones to the query and thus less informative. The insufficiency of training samples both in quantities and varieties constrains the generalization ability of the classifier significantly. In this paper, we propose a novel relevance feedback approach, which aims to collect more representative samples and hence improve the performance of classifier. Image labeling and classifier training are conducted in two complementary image feature spaces. Since the samples distribute differently in two spaces, the positive samples may be more informative in one feature space than in another. The two complementary feature spaces are alternated iteratively during the feedback process. To choose appropriate complementary feature spaces, we present two methods to measure the complementarities between two feature spaces quantitatively. Our experimental result on 10,000 images indicates that the proposed feedback approach significantly improves image retrieval performance.     

Incorporating multiple distance spaces in optimum-path forest classification to improve feedback-based learning

In content-based image retrieval (CBIR) using feedback-based learning, the user marks the relevance of returned images and the system learns how to return more relevant images in a next iteration. In this learning process, image comparison may be based on distinct distance spaces due to multiple visual content representations. This work improves the retrieval process by incorporating multiple distance spaces in a recent method based on optimum-path forest (OPF) classification. For a given training set with relevant and irrelevant images, an optimization algorithm finds the best distance function to compare images as a combination of their distances according to different representations. Two optimization techniques are evaluated: a multi-scale parameter search (MSPS), never used before for CBIR, and a genetic programming (GP) algorithm. The combined distance function is used to project an OPF classifier and to rank images classified as relevant for the next iteration. The ranking process takes into account relevant and irrelevant representatives, previously found by the OPF classifier. Experiments show the advantages in effectiveness of the proposed approach with both optimization techniques over the same approach with single distance space and over another state-of-the-art method based on multiple distance spaces.     

Neighborhood preserving regression for image retrieval

In content-based image retrieval (CBIR), relevance feedback has been proven to be a powerful tool for bridging the gap between low level visual features and high level semantic concepts. Traditionally, relevance feedback driven CBIR is often considered as a supervised learning problem where the user provided feedbacks are used to learn a distance metric or classification function. However, CBIR is intrinsically a semi-supervised learning problem in which the testing samples (images in the database) are present during the learning process. Moreover, when there are no sufficient feedbacks, these methods may suffer from the overfitting problem. In this paper, we propose a novel neighborhood preserving regression algorithm which makes efficient use of both labeled and unlabeled images. By using the unlabeled images, the geometrical structure of the image space can be incorporated into the learning system through a regularizer. Specifically, from all the functions which minimize the empirical loss on the labeled images, we select the one which best preserves the local neighborhood structure of the image space. In this way, our method can obtain a regression function which respects both semantic and geometrical structures of the image database. We present experimental evidence suggesting that our algorithm is able to use unlabeled data effectively for image retrieval.     

Boosting Image Retrieval

We present an approach for image retrieval using a very large number of highly selective features and efficient learning of queries. Our approach is predicated on the assumption that each image is generated by a sparse set of visual “causes” and that images which are visually similar share causes. We propose a mechanism for computing a very large number of highly selective features which capture some aspects of this causal structure (in our implementation there are over 46,000 highly selective features). At query time a user selects a few example images, and the AdaBoost algorithm is used to learn a classification function which depends on a small number of the most appropriate features. This yields a highly efficient classification function. In addition we show that the AdaBoost framework provides a natural mechanism for the incorporation of relevance feedback. Finally we show results on a wide variety of image queries.     

Complementary relevance feedback-based content-based image retrieval

We propose a complementary relevance feedback-based content-based image retrieval (CBIR) system. This system exploits the synergism between short-term and long-term learning techniques to improve the retrieval performance. Specifically, we construct an adaptive semantic repository in long-term learning to store retrieval patterns of historical query sessions. We then extract high-level semantic features from the semantic repository and seamlessly integrate low-level visual features and high-level semantic features in short-term learning to effectively represent the query in a single retrieval session. The high-level semantic features are dynamically updated based on users’ query concept and therefore represent the image’s semantic concept more accurately. Our extensive experimental results demonstrate that the proposed system outperforms its seven state-of-the-art peer systems in terms of retrieval precision and storage space on a large scale imagery database.     

Multi-Class Learning by Smoothed Boosting

AdaBoost.OC has been shown to be an effective method in boosting “weak” binary classifiers for multi-class learning. It employs the Error-Correcting Output Code (ECOC) method to convert a multi-class learning problem into a set of binary classification problems, and applies the AdaBoost algorithm to solve them efficiently. One of the main drawbacks with the AdaBoost.OC algorithm is that it is sensitive to the noisy examples and tends to overfit training examples when they are noisy. In this paper, we propose a new boosting algorithm, named “MSmoothBoost”, which introduces a smoothing mechanism into the boosting procedure to explicitly address the overfitting problem with AdaBoost.OC. We proved the bounds for both the empirical training error and the marginal training error of the proposed boosting algorithm. Empirical studies with seven UCI datasets and one real-world application have indicated that the proposed boosting algorithm is more robust and effective than the AdaBoost.OC algorithm for multi-class learning. Editor: Nicolo Cesa-Bianchi     

基于PCA与主动学习AdaBoost的频谱感知算法

针对认知网络中各低信噪比环境下主用户信号检测率偏低的问题,提出一种基于主成分分析和主动学习AdaBoost的主用户信号频谱感知算法.该算法首先采用主成分分析算法对信号特征参数进行提取,获得信号的主成分,之后利用主动学习算法通过多次迭代抽样,获取有利于提高分类性能的样本,并对AdaBoost分类器进行训练,最后利用训练完成的AdaBoost分类器对待测信号进行分类检测.仿真实验表明,在各低信噪比情况下与ANN和MME算法相比较,所提算法具有较高的分类感知性能,有效地实现了对主用户信号的频谱感知.     

基于内容的图像检索中SVM和Boosting方法集成应用

提出一种适用于图像内容检索的AdaBoostSVM算法。算法思想是采用支持向量机(SVM)作为AdaBoost算法的分量分类器;基于相关反馈检索机制,通过增加重要样本来模拟AdaBoost算法的权重调整方法。在包含2000幅图像的数据库中进行了检索实验,结果表明AdaBoostSVM算法能有效提高系统的检索性能。     

A novel relevance feedback method for CBIR

In this paper, we address the challenge about insufficiency of training set and limited feedback information in each relevance feedback (RF) round during the process of content based image retrieval (CBIR). We propose a novel active learning scheme to utilize the labeled and unlabeled images to build the initial Support Vector Machine (SVM) classifier for image retrieving. In our framework, two main components, a pseudo-label strategy and an improved active learning selection method, are included. Moreover, a feature subspace partition algorithm is proposed to model the retrieval target from users by the analysis from relevance labeled images. Experimental results demonstrate the superiority of the proposed method on a range of databases with respect to the retrieval accuracy.     

Fast shared boosting for large-scale concept detection

Visual concept detection consists in assigning labels to an image or keyframe based on its semantic content. Visual concepts are usually learned from an annotated image or video database with a machine learning algorithm, posing this problem as a multiclass supervised learning task. Some practical issues appear when the number of concept grows, in particular in terms of available memory and computing time, both for learning and testing. To cope with these issues, we propose to use a multiclass boosting algorithm with feature sharing and reduce its computational complexity with a set of efficient improvements. For this purpose, we explore a limited part of the possible parameter space, by adequately injecting randomness into the crucial steps of our algorithm. This makes our algorithm able to handle a problem of classification with many classes in a reasonable time, thanks to a linear complexity with regards to the number of concepts considered as well as the number of feature and their size. The relevance of our algorithm is evaluated in the context of information retrieval, on the benchmark proposed into the ImageCLEF international evaluation campaign and shows competitive results.     

基于多分类AdaBoost改进算法的TEE标准切面分类

针对超声图像样本冗余、不同标准切面因疾病导致的高度相似性、感兴趣区域定位不准确问题,提出一种结合特征袋（BOF）特征、主动学习方法和多分类AdaBoost改进算法的经食管超声心动图（TEE）标准切面分类方法。首先采用BOF方法对超声图像进行描述;然后采用主动学习方法选择对分类器最有价值的样本作为训练集;最后,在AdaBoost算法对弱分类器的迭代训练中,根据临时强分类器的分类情况调整样本更新规则,实现对多分类AdaBoost算法的改进和TEE标准切面的分类。在TEE数据集和三个UCI数据集上的实验表明,相比AdaBoost.SAMME算法、多分类支持向量机（SVM）算法、BP神经网络和AdaBoost.M2算法,所提算法在各个数据集上的G-mean指标、整体分类准确率和大多数类别分类准确率都有不同程度的提升,且比较难分的类别分类准确率提升最为显著。实验结果表明,在包含类间相似样本的数据集上,分类器的性能有显著提升。     

面向草图检索的小样本增量有偏学习算法

为了解决草图检索相关反馈中小样本训练、数据不对称及实时性要求这3个难点问题,提出了一种小样本增量有偏学习算法.该算法将主动式学习、有偏分类和增量学习结合起来,对相关反馈过程中的小样本有偏学习问题进行建模.其中,主动式学习通过不确定性采样,选择最佳的用户标注样本,实现有限训练样本条件下分类器泛化能力的最大化;有偏分类通过构造超球面区别对待正例和反例,准确挖掘用户目标类别;每次反馈循环中新加入的样本则用于分类器的增量学习,在减少分类器训练时间的同时积累样本信息,进一步缓解小样本问题.实验结果表明,该算法可以有效地改善草图检索性能,也适用于图像检索和三维模型检索等应用领域.     

Fuzzy SVM for content-based image retrieval: a pseudo-label support vector machine framework

Conventional relevance feedback in content-based image retrieval (CBIR) systems uses only the labeled images for learning. Image labeling, however, is a time-consuming task and users are often unwilling to label too many images during the feedback process. This gives rise to the small sample problem where learning from a small number of training samples restricts the retrieval performance. To address this problem, we propose a technique based on the concept of pseudo-labeling in order to enlarge the training data set. As the name implies, a pseudo-labeled image is an image not labeled explicitly by the users, but estimated using a fuzzy rule. Therefore, it contains a certain degree of uncertainty or fuzziness in its class information. Fuzzy support vector machine (FSVM), an extended version of SVM, takes into account the fuzzy nature of some training samples during its training. In order to exploit the advantages of pseudo-labeling, active learning and the structure of FSVM, we develop a unified framework called pseudo-label fuzzy support vector machine (PLFSVM) to perform content-based image retrieval. Experimental results based on a database of 10,000 images demonstrate the effectiveness of the proposed method.     

An improved distance-based relevance feedback strategy for image retrieval

Most CBIR (content based image retrieval) systems use relevance feedback as a mechanism to improve retrieval results. NN (nearest neighbor) approaches provide an efficient method to compute relevance scores, by using estimated densities of relevant and non-relevant samples in a particular feature space. In this paper, particularities of the CBIR problem are exploited to propose an improved relevance feedback algorithm based on the NN approach. The resulting method has been tested in a number of different situations and compared to the standard NN approach and other existing relevance feedback mechanisms. Experimental results evidence significant improvements in most cases.     

Batch Mode Active Learning with Applications to Text Categorization and Image Retrieval

Most machine learning tasks in data classification and information retrieval require manually labeled data examples in the training stage. The goal of active learning is to select the most informative examples for manual labeling in these learning tasks. Most of the previous studies in active learning have focused on selecting a single unlabeled example in each iteration. This could be inefficient, since the classification model has to be retrained for every acquired labeled example. It is also inappropriate for the setup of information retrieval tasks where the user's relevance feedback is often provided for the top K retrieved items. In this paper, we present a framework for batch mode active learning, which selects a number of informative examples for manual labeling in each iteration. The key feature of batch mode active learning is to reduce the redundancy among the selected examples such that each example provides unique information for model updating. To this end, we employ the Fisher information matrix as the measurement of model uncertainty, and choose the set of unlabeled examples that can efficiently reduce the Fisher information of the classification model. We apply our batch mode active learning framework to both text categorization and image retrieval. Promising results show that our algorithms are significantly more effective than the active learning approaches that select unlabeled examples based only on their informativeness for the classification model.     

Image retrieval based on augmented relational graph representation

The “semantic gap” problem is one of the main difficulties in image retrieval tasks. Semi-supervised learning, typically integrated with the relevance feedback techniques, is an effective method to narrow down the semantic gap. However, in semi-supervised learning, the amount of unlabeled data is usually much greater than that of labeled data. Therefore, the performance of a semi-supervised learning algorithm relies heavily on its effectiveness of using the relationships between the labeled and unlabeled data. This paper proposes a novel algorithm to better explore those relationships by augmenting the relational graph representation built on the entire data set, expected to increase the intra-class weights while decreasing the inter-class weights and linking the potential intra-class data. The augmented relational matrix can be directly used in any semi-supervised learning algorithms. The experimental results in a range of feedback-based image retrieval tasks show that the proposed algorithm not only achieves good generality, but also outperforms other algorithms in the same semi-supervised learning framework.     

基于多尺度生成对抗网络的遮挡行人重识别方法

行人重识别是指在多个非重叠摄像头拍摄的场景下,给定一幅查询行人图像,从大规模行人图像库中检索出具有相同身份的行人图像,是一类特殊的图像检索任务.随着深度学习的不断发展,行人重识别方法的性能得到了显著提升.但是行人重识别在实际应用中经常遭遇遮挡问题（例如背景遮挡、行人互相遮挡等）.由于遮挡图像不仅丢失了部分目标信息,而且引入了额外的干扰,使得现有方法往往难以学习到鲁棒的特征表示,从而导致识别性能严重下降.最近,生成对抗网络在各类计算机视觉任务上展现出强大的图像生成能力.受到生成对抗网络的启发,提出了一种基于多尺度生成对抗网络的遮挡行人重识别方法.首先,利用成对的遮挡图像和非遮挡图像训练一个多尺度生成器和一个判别器.多尺度生成器能够对随机遮挡区域进行去遮挡操作,生成高质量的重构图;而判别器能够区分输入图像是真实图像还是生成图像.其次,利用训练好的多尺度生成器,生成去除随机遮挡的训练图像,添加到原始训练图像集,用于增加训练样本的多样性.最后,基于此扩充训练图像集,训练分类识别模型,有效地提高模型在测试图像集上的泛化性.在多个有挑战性的行人重识别数据集上的实验结果,验证了所提出方法的有效性.     

基于多部位多示例学习的人体检测

基于部位的检测方法能处理多姿态及部分遮挡的人体检测,多示例学习能有效处理图像的多义性,被广泛应用于图像检索与场景理解中。文中提出一种基于多示例学习的多部位人体检测方法。首先,根据人体生理结构将图像分割成若干区域,每个区域包含多个示例,利用AdaBoost多示例学习算法来训练部位检测器。然后利用各部位检测器对训练样本进行测试得到其响应值,从而将训练样本转化为部位响应值组成的特征向量。再用SVM方法对这些向量进行学习,得到最终的部位组合分类器。在INRIA数据集上的实验结果表明该方法能改进单示例学习的检测性能,同时评价3种不同的部位划分及其对检测性能的影响。     

A self-constructing cascade classifier with AdaBoost and SVM for pedestriandetection

In this paper, we propose a cascade classifier combining AdaBoost and support vector machine, and applied this to pedestrian detection. The pedestrian detection involved using a window of fixed size to extract the candidate region from left to right and top to bottom of the image, and performing feature extractions on the candidate region. Finally, our proposed cascade classifier completed the classification of the candidate region. The cascade-AdaBoost classifier has been successfully used in pedestrian detection. We have improved the initial setting method for the weights of the training samples in the AdaBoost classifier, so that the selected weak classifier would be able to focus on a higher detection rate other than accuracy. The proposed cascade classifier can automatically select the AdaBoost classifier or SVM to construct a cascade classifier according to the training samples, so as to effectively improve classification performance and reduce training time. In order to verify our proposed method, we have used our extracted database of pedestrian training samples, PETs database, INRIA database and MIT database. This completed the pedestrian detection experiment whose result was compared to those of the cascade-AdaBoost classifier and support vector machine. The result of the experiment showed that in a simple environment involving campus experimental image and PETs database, both our cascade classifier and other classifiers can attain good results, while in a complicated environment involving INRA and MIT database experiments, our cascade classifier had better results than those of other classifiers.     

一种新的基于SVM和主动学习的图像检索方法

在基于内容的图像检索中,支持向量机（SVM）能够很好地解决小样本问题,而主动学习算法则可以根据学习进程主动选择最佳的样本进行学习,大幅度缩短训练时间,提高分类算法效率。为使图像检索更加快速、高效,提出一种新的基于SVM和主动学习的图像检索方法。该方法根据SVM构造分类器,通过“V”型删除法快速缩减样本集,同时通过最优选择法从缩减样本集中选取最优的样本作为训练样本,最终构造出不仅信息度大而且冗余度低的最优训练样本集,从而训练出更好的SVM分类器,得到更高的检索效率。实验结果表明,与传统的SVM主动学习的图像检索方法相比,该方法能够较大幅度提高检索性能。