Multiproxies Adaptive Distribution Loss with Weakly Supervised Feature Aggregation for Fine-Grained Retrieval

Fine-grained retrieval is one of the complex problems in computer vision. Compared with general content-based image retrieval, fine-grained image retrieval faces more difficult challenges. In fine-grained image retrieval tasks, all classes belong to a subclass of a meta-class, so there will be small interclass variance and large intraclass variance. In order to solve this problem, in this paper, we propose a fine-grained retrieval method to improve loss and feature aggregation, which can achieve better retrieval results under a unified framework. Firstly, we propose a novel multiproxies adaptive distribution loss which can better characterize the intraclass variations and the degree of dispersion of each cluster center. Secondly, we propose a weakly supervised feature aggregation method based on channel weighting, which distinguishes the importance of different feature channels to obtain more representative image feature descriptors. We verify the performance of our proposed method on the universal benchmark datasets such as CUB200-2011 and Stanford Dog. Higher Recall@K demonstrates the advantage of our proposed method over the state of the art.

IMAGE MEDIA AND THE RESULTS OF LEARNING OUTCOMES OF KONTRASTIVE KULTURKUNDE COURSES OF THE STUDENTS OF GERMAN LANGUAGE EDUCATION STUDY PROGRAM, FKIP PATTIMURA UNIVERSITY, AMBON

The purpose of this study is to determine the influence of the use of image media on the learning outcomes of students of the German Language Education Study Program in Kontrastive Kulturkunde courses that use image media and those that do not use images. The use of image media applied in this study follows what Macaire and Hosch (2004: 67) have suggested about the function of images, that images can reactivate students' initial knowledge (Vorwissen der Schülerinnen und Schüler), show similarities/authenticity (Authentizität) and indicate actual situations (Aktualität), a communicative perspective image of an object and a representative image, which is a visual that depicts the state of nature and people's lives. (Perspektive und Repräsentativität) and which have similarities or differ with their own culture (Nähe oder Ferne zur eigenen Kultur). This research was motivated by the low learning outcomes of students in Kontrastive Kulturkunde courses. The type of this research used is experiment with one group pretest - posttest design with t-test analysis technique with significance level α = 0.05. Testing the effectiveness of this study is with the use of image media. The target of this study was 20 students of semester 4, German Language Education Study Program FKIP Pattimura University Ambon in Kontrastive Kulturkunde courses. The result of the calculation appears that, tcount > ttable=1,729 which is 11,103 > 1,729. Based on the results of the calculation it can be said that, there is a noticeable difference between the value after and before the application of the image media

Deep Unified Cross-Modality Hashing by Pairwise Data Alignment

With the increasing amount of multimedia data, cross-modality hashing has made great progress as it achieves sub-linear search time and low memory space. However, due to the huge discrepancy between different modalities, most existing cross-modality hashing methods cannot learn unified hash codes and functions for modalities at the same time. The gap between separated hash codes and functions further leads to bad search performance. In this paper, to address the issues above, we propose a novel end-to-end Deep Unified Cross-Modality Hashing method named DUCMH, which is able to jointly learn unified hash codes and unified hash functions by alternate learning and data alignment. Specifically, to reduce the discrepancy between image and text modalities, DUCMH utilizes data alignment to learn an auxiliary image to text mapping under the supervision of image-text pairs. For text data, hash codes can be obtained by unified hash functions, while for image data, DUCMH first maps images to texts by the auxiliary mapping, and then uses the mapped texts to obtain hash codes. DUCMH utilizes alternate learning to update unified hash codes and functions. Extensive experiments on three representative image-text datasets demonstrate the superiority of our DUCMH over several state-of-the-art cross-modality hashing methods.

Doorway to Europe: migration and its impact on island tourism

Lampedusa is today best known in relation to migration. The island has confronted one migration crisis after the other for the past three decades, resulting in extensive media coverage. Whereas Lampedusa, a small island in the central Mediterranean region, has an economy mainly based on tourism, its name remains associated with migration, which is believed to negatively impact the island’s image and the performance of its tourism sector. On the other hand, migration has to some extent, put Lampedusa on the map, helping the island gain popularity. The island’s existing tourism model is based on sun, sand, and sea (3S) and is made attractive by its beaches, one of which has ranked as the best beach in the world, as well as by marketing efforts presenting Lampedusa as 'the Caribbean island of the Mediterranean'. However, migration and 3S are camouflaging other resources that are key to the island’s image. Lampedusa has a terrestrial nature reserve as well as a marine protected area that is home to several charismatic marine species, making it an ideal ecotourism destination. These resources can be used not only to depict a more representative image of the island but also to develop a sustainable tourism model that is suitable to a small island.

MRI-Based Classification Scheme for Degenerative Cervical Spinal Stenosis (DCSS Classification)

Background: The severity of Degenerative Cervical Spinal Stenosis (DCSS) is currently assessed by determining the sagittal diameter of the spinal canal, the degree of narrowing, and the level of stenosis using magnetic resonance imaging (MRI). The aim of our study was to develop and evaluate an optimized, MRI-based classification of DCSS to support treatment decision-making. Methods: We analyzed preoperative MRI scans of 75 patients to determine key morphological features of DCSS. Based on the features found, a classification was proposed and tested on ten representative image sets by 53 neurosurgeons to determine practicability and inter-observer reliability. Results: DCSS extended over one (13 and 17 % of pat.), two (23 and 31 % of pat.) or three (39 and 52 % of pat.). Thickening of the ligaments was observed in 71%, scoliosis in 3% and grade I spondylolisthesis in 11% of the patients. The DCSS classification has three main Types (A, B, C), which differ in the extent of the stenoses and thickening of the ligaments. Each main class has 2-3 subtypes depending on the position of the stenosis (lateral/medial) and the number of stenotic segments. An additional suffix indicates the presence of concomitant pathology (spondylolisthesis, scoliosis). 26 (49 %) neurosurgeons rated the classification as useful. For the main classes, interobserver reliability was fair (k=0.23). For subtype and concomitant pathology, interobserver reliability was low (k=0.14). Conclusions: In the study a new classification system for degenerative spinal stenoses was developed, which combines anatomical and clinical defect characteristics. Deviations in the classification result from a strict subdivision, so that adjustments are necessary.

Komunistiskās partijas kontroles mehānisms Latvijas Padomju rakstnieku savienībā: Žaņa Grīvas piemērs

It was in the interests of the Communist Party to create a representative image of Latvian Soviet writers, which would represent the interests of the party and at the same time oversee the course of literary life in the Latvian Soviet Writers’ Union. Such was the writer Žanis Grīva in the Latvian Soviet literary process. The influential positions in the Soviet nomenclature gave him power in the creative environment and created opportunities to monitor the implementation of the Communist Party’s course. The article aims to put forward the personality of Žanis Grīva in the context of the research of the Latvian soviet literary process and the Latvian Soviet Writers’ Union, proposing several issues to be further researched and developed in the future. The article is based on the documents of the Latvian Soviet Writers’ Union and the Communist Party local organization of the Latvian Soviet Writers’ Union, and the personal file of Žanis Grīva in the Central Committee of the Communist Party of Latvia, and documents of the Žanis Grīva collection that are available at the Latvian State Archive of the National Archives of Latvia, as well as Žanis Grīva’s personal documents regarding his life and professional activities that are available at the Aleksejs Apinis Rare Books and Manuscripts Reading Room at the National Library of Latvia. Memories of contemporaries were also investigated. It is concluded that the role of Žanis Grīva in the Latvian soviet literary process and the Latvian Soviet Writers’ Union is political and purposefully constructed by the Communist Party, and has little to do with literature and literary talents. It can be assumed that Žanis Grīva has negatively affected the creative activity of some members of the Writers’ Union, such as Gunārs Priede.

City, history, and memory: from a destroyed environment to a constructed one. A case study of Natividade da Serra, state of São Paulo, Brazil

Between 1973 and 1974, the city of Natividade da Serra, state of São Paulo, was relocated in order to provide space to construct a hydroelectric dam for power generation. Based on the idea that the relationship between society and physical space acts to shape and influence the culture of a society, we conducted this study with the principal objective of analyzing the social impacts on the local community by highlighting the cultural transformations provoked by the restructuring of a space that was constructed over the period of a century. By conducting a qualitative analysis of data from the document repository of the City Hall and Parish of Our Lady of the Nativity (Nossa Senhora da Natividade), and from oral sources (descriptions collected through oral history), we elaborated a representative image of daily life in the city as it existed before the restructuring. This reconstruction started with the process of destruction and then proceeded to describe the reorganization of the community in the new space, as well as detailing perceptions of the city as a place that has passed through multiple historical experiences. By relating the data obtained from the written and oral sources to the historical context, we attempted to reveal the political, cultural, and ideological dimensions involved in this process of manipulation of inhabited space. We conclude that the innumerous implications caused by the abrupt transformation of this space has altered the customs and forms of sociability of the population. The analysis of the reasons and motivations for the disappearance of the city suggests that, even if only implicitly, this cultural transformation was desired by those who were in power in Brazil at that moment in time.

Development of a new cultural design process using Kansei engineering and fuzzy techniques

Purpose Belief in Mazu has a crucial cultural status in Taiwan and the coastal area of Fujian, China. The design and manufacture of apparel and accessories to be placed on statues of the deity are also considered a sacred and critical part of the religion’s cultural and artistic inheritance. The crown hat of Mazu is one of the most essential elements of the deity’s apparel. The paper aims to discuss these issues. Design/methodology/approach This study explored the styles of Mazu crown hats using Kansei engineering (KE). People generally use adjectives words to provide aesthetic evaluations. Fuzzy theory is suitable for processing linguistic problems that include vagueness, thereby providing a reasonable method of quantifying such aesthetic evaluations. Therefore, this study first established a fuzzy positioning model (FPM) of word evaluations for analysis. Factor analysis was used to obtain representative image adjectives that represented Mazu’s image. Fuzzy analysis methods were then employed to rank the various image adjectives through evaluation words and to determine the differences between adjectives. Finally, on the basis of image analysis results and expert suggestions, the crown hat was redesigned and its suitability verified. Findings Four results were obtained. First, four image adjectives appropriate for representing Mazu’s image were identified, of which “noble and kind” is the most suitable. Second, fuzzy analysis was found to successfully rank style images. Third, the crown hat style and design characteristics suitable for Mazu were acquired. Fourth, the verification demonstrated that the redesign effectively enhanced the perceived image of the crown hat design. Originality/value This study employed KE to improve the design of a Mazu crown hat. The proposed FPM can aid the development of cultural and creative design.