Кластеризация видеопоследовательностей в системах видеонаблюдения на основе сверточных нейронных сетей
The article is devoted to pattern recognition task with the database containing small number of samples per class. By mapping of local continuous feature vectors to a discrete range, this problem is reduced to statistical classification of a set of discrete finite patterns. It is demonstrated that Bayesian decision under the assumption that probability distributions can be estimated using the Parzen kernel and the Gaussian window with a fixed variance for all the classes, implemented in the PNN, is not optimal in the classification of a set of patterns. We presented here the novel modification of the PNN with homogeneity testing which gives an optimal solution of the latter task under the same assumption about probability densities. By exploiting the discrete nature of patterns our modification prevents the well-known drawbacks of the memory-based approach implemented in both the PNN and the PNN with homogeneity testing, namely, low classification speed and high requirements to the memory usage. Our modification only requires the storage and processing of the histograms of input and training samples. We present the results of an experimental study in two practically important tasks: 1) the problem of Russian text authorship attribution with character n-grams features; and 2) face recognition with well-known datasets (AT&T, FERET and JAFFE) and comparison of color- and gradient-orientation histograms. Our results support the statement that the proposed network provides better accuracy (1-7%) and is much more resistant to change of the smoothing parameter of Gaussian kernel function in comparison with the original PNN.
This state-of-the-art survey is dedicated to the memory of Emmanuil Markovich Braverman (1931-1977), a pioneer in developing the machine learning theory. The 12 revised full papers and 4 short papers included in this volume were presented at the conference "Braverman Readings in Machine Learning: Key Ideas from Inception to Current State" held in Boston, MA, USA, in April 2017, commemorating the 40th anniversary of Emmanuil Braverman's decease. The papers present an overview of some of Braverman's ideas and approaches. The collection is divided in three parts. The first part bridges the past and the present. Its main contents relate to the concept of kernel function and its application to signal and image analysis as well as clustering. The second part presents a set of extensions of Braverman's work to issues of current interest both in theory and applications of machine learning. The third part includes short essays by a friend, a student, and a colleague.
The sixteen-volume set comprising the LNCS volumes 11205-11220 constitutes the refereed proceedings of the 15th European Conference on Computer Vision, ECCV 2018, held in Munich, Germany, in September 2018.
The 776 revised papers presented were carefully reviewed and selected from 2439 submissions. The papers are organized in topical sections on learning for vision; computational photography; human analysis; human sensing; stereo and reconstruction; optimization; matching and recognition; video attention; and poster sessions.
This is a textbook in data analysis. Its contents are heavily influenced by the idea that data analysis should help in enhancing and augmenting knowledge of the domain as represented by the concepts and statements of relation between them. According to this view, two main pathways for data analysis are summarization, for developing and augmenting concepts, and correlation, for enhancing and establishing relations. Visualization, in this context, is a way of presenting results in a cognitively comfortable way. The term summarization is understood quite broadly here to embrace not only simple summaries like totals and means, but also more complex summaries such as the principal components of a set of features or cluster structures in a set of entities.
The material presented in this perspective makes a unique mix of subjects from the fields of statistical data analysis, data mining, and computational intelligence, which follow different systems of presentation.
The paper describes the results of an experimental study of topic models applied to the task of single-word term extraction. The experiments encompass several probabilistic and non-probabilistic topic models and demonstrate that topic information improves the quality of term extraction, as well as NMF with KL-divergence minimization is the best among the models under study.
Technology mining (TM) helps to acquire intelligence about the evolution of research and development (R&D), technologies, products, and markets for various STI areas and what is likely to emerge in the future by identifying trends. The present chapter introduces a methodology for the identification of trends through a combination of “thematic clustering” based on the co-occurrence of terms, and “dynamic term clustering” based on the correlation of their dynamics across time. In this way, it is possible to identify and distinguish four patterns in the evolution of terms, which eventually lead to (i) weak signals of future trends, as well as (ii) emerging, (iii) maturing, and (iv) declining trends. Key trends identified are then further analyzed by looking at the semantic connections between terms identified through TM. This helps to understand the context and further features of the trend. The proposed approach is demonstrated in the field photonics as an emerging technology with a number of potential application areas.
The goal of the study is to increase the computation efficiency of the face recognition that uses feature vectors to describe facial images on photos and videos. These high-dimensional feature vectors are nowadays produced by convolutional neural networks. The methods to aggregate the features generated for each video frame are used to process the video sequences. A novel hierarchical recognition algorithm is proposed. In contrast to known approaches our algorithm seeks the nearest neighbors only among reference images of most reliable classes selected at the preceding stage to carry out the sequential analysis of a more detailed description level (with a greater dimensionality of the feature vector). At each stage principal components are compared, the number of the components being chosen according to a given portion of explained variations. Datasets like Labeled Faces in the Wild, YouTubeFaces, IAPRA, Jenus Benchmark C and different neural-net face descriptors are used to compare the algorithm with other methods. In contrast with the conventional nearest-neighbor method, the proposed approach is shown to allow a 2- to 16-times reduction of the classifier running time.