Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Pattern Recognition and Machine Learning

“Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks”の学習報告

From the Publisher: 
The accessible presentation of this book gives both a general view of the entire computer vision enterprise and also offers sufficient detail to be able to build useful applications. Users learn techniques that have proven to be useful by first-hand experience and a wide range of mathematical methods. A CD-ROM with every copy of the text contains source code for programming practice, color images, and illustrative movies. Comprehensive and up-to-date, this book includes essential topics that either reflect practical significance or are of theoretical importance. Topics are discussed in substantial and increasing depth. Application surveys describe numerous important application areas such as image based rendering and digital libraries. Many important algorithms broken down and illustrated in pseudo code. Appropriate for use by engineers as a comprehensive reference to the computer vision enterprise.

Computer vision : a modern approach = 计算机视觉 : 一种现代的方法

Color image segmentation: advances and prospects

When considering person re-identification (re-ID) as a retrieval process, re-ranking is a critical step to improve its accuracy. Yet in the re-ID community, limited effort has been devoted to re-ranking, especially those fully automatic, unsupervised solutions. In this paper, we propose a k-reciprocal encoding method to re-rank the re-ID results. Our hypothesis is that if a gallery image is similar to the probe in the k-reciprocal nearest neighbors, it is more likely to be a true match. Specifically, given an image, a k-reciprocal feature is calculated by encoding its k-reciprocal nearest neighbors into a single vector, which is used for re-ranking under the Jaccard distance. The final distance is computed as the combination of the original distance and the Jaccard distance. Our re-ranking method does not require any human interaction or any labeled data, so it is applicable to large-scale datasets. Experiments on the large-scale Market-1501, CUHK03, MARS, and PRW datasets confirm the effectiveness of our method.

/pdf/re-ranking-person-re-identification-with-k-reciprocal-3erzmilgv1.pdf

Re-ranking Person Re-identification with k-Reciprocal Encoding

We consider concept representation in the framework of conceptual spaces. We propose the introduction of fuzzy sets to express the entries of the concept (co-occurrence) matrix as fuzzy probabilities. Initial experimental results indicate that the proposed approach can support a much more sophisticated and useful concept representation.

Fuzzy sets in concept representation using conceptual spaces

The development of technology concerning semiconductors and optical fibers now makes image transmission cheaper and more practical. This paper describes an error correcting coding method for binary images of line drawings. First the bilevel images of line drawings are represented by legal patterns, which are defined with a 3 × 3-pixel unit mesh (window) and employ line constraints. The resulting image is a picture represented by a legal symbol connection (LSC-picture). This picture includes certain connecting relations among the neighboring legal patterns, which are considered to represent explicitly the redundancy of the original image. Next we encode the picture using a legal pattern as a coding unit. The information source model used for encoding is a simple Markov model. We use line-to-line dependencies for the purpose of error detection and correction. When a transmission error is detected by checking the relations, the decoding process generates a tree called an “error correcting tree.” The problem of correcting the transmission error is considered as one of how to search the error correcting tree. The simulation results of this method show that almost all the transmission errors can be corrected if we incorporate a little redundancy of the code words themselves, as in an ordinary method. At the same time, the compression ratio is as high as that of the MR coding method. For image transmission in the future we believe that a description should be considered which can make explicit the redundancy of the image itself as much as possible and that the redundancy should be utilized not only for compressing the amount of data but also for correcting transmission errors.

An error detecting and correcting coding method for bilevel images of line drawings using the redundancy of the images

A method of tracking the context in man-machine communication is proposed. The “tracking the context” is defined as locating the place of context switching and dividing the user's command sequence into chunks, each of which corresponds to one work. For tracking the context, “a syntactic-context” is introduced. A syntactic-context is a working set of words in the command lines a user enters. To locate the place of the syntactic-context switching, word-kind ratio and window contribution are defined. It is shown that the hit rate of command prediction is improved by using these indicators, which shows the effectiveness of the proposed method.

Tracking the context in man-machine communication

In an image retrieval system, accurately extracting the content of the image which users want to search for based on an image in their heads is vital. As a result, many attempts have been made to use an image similar to the sought-for image as a retrieval key and thereby to increase the freedom of expression for users' images. Because of the high variability of image characteristics, however, the vagueness of the content which indicates the retrieval key itself rises compared to conventional methods that use keywords or other linguistic resources for the retrieval key. Therefore, in this research the authors consider these problems by reviewing image retrieval from the perspective of image communications between users and the system, and then propose an interactive mechanism which limits the increase in vagueness while using an image as the retrieval key. The authors then perform retrievals by having users revise the retrieval key based on this mechanism. The authors create their interactive mechanism in a real image retrieval system and confirm its validity through experiments. © 2001 Scripta Technica, Syst Comp Jpn, 32(13): 41–51, 2001

An interactive image retrieval system with self-explanation function

We propose a new coding method by applying slant transform (SLT) to gradation areas instead of discrete cosine transform (DCT). A contour in a gradation area, which is often caused by the coding process for high compression, is one of the typical artifact noises for subjective evaluation. Since the pixel value changes linearly in a gradation area, DCT is not always suitable for its transformation. In our proposed method, although only traditional DCT is utilized in the encoding process to avoid increasing computational complexity, the inverse SLT (ISLT) instead of IDCT is performed in the area that is estimated as a gradation area in accordance with the distribution of DCT coefficients in the decoding process. The coding block size of the gradation area is easily expanded by simple operations using the hierarchy of SLT. In accordance with our experimental analysis, the proposed scheme has been performed to clarify the improvement of quality in gradation areas.

Michihiko Minoh

Papers

Fuzzy sets in concept representation using conceptual spaces

An error detecting and correcting coding method for bilevel images of line drawings using the redundancy of the images

Tracking the context in man-machine communication

An interactive image retrieval system with self-explanation function

High Quality Image Coding Method for Gradation Areas by Using Slant Transform