Efficient Image and Video Representations for Retrieval

Image (Video) retrieval is an interesting problem of retrieving images (videos) similar to the query. Images (Videos) are represented in an input (feature) space and similar images (videos) are obtained by finding nearest neighbors in the input representation space. Numerous input representations both in real valued and binary space have been proposed for conducting faster retrieval. In this thesis, we present techniques that obtain improved input representations for retrieval in both supervised and unsupervised settings for images and videos. Supervised retrieval is a well known problem of retrieving same class images of the query. We address the practical aspects of achieving faster retrieval with binary codes as input representations for the supervised setting in the first part, where binary codes are used as addresses into hash tables. In practice, using binary codes as addresses does not guarantee fast retrieval, as similar images are not mapped to the same binary code (address). We address this problem by presenting an efficient supervised hashing (binary encoding) method that aims to explicitly map all the images of the same class ideally to a unique binary code. We refer to the binary codes of the images as `Semantic Binary Codes' and the unique code for all same class images as `Class Binary Code'. We also propose a new class­ based Hamming metric that dramatically reduces the retrieval times for larger databases, where only hamming distance is computed to the class binary codes. We also propose a Deep semantic binary code model, by replacing the output layer of a popular convolutional Neural Network (AlexNet) with the class binary codes and show that the hashing functions learned in this way outperforms the state­ of ­the art, and at the same time provide fast retrieval times. In the second part, we also address the problem of supervised retrieval by taking into account the relationship between classes. For a given query image, we want to retrieve images that preserve the relative order i.e. we want to retrieve all same class images first and then, the related classes images before different class images. We learn such relationship aware binary codes by minimizing the similarity between inner product of the binary codes and the similarity between the classes. We calculate the similarity between classes using output embedding vectors, which are vector representations of classes. Our method deviates from the other supervised binary encoding schemes as it is the first to use output embeddings for learning hashing functions. We also introduce new performance metrics that take into account the related class retrieval results and show significant gains over the state­ of­ the art. High Dimensional descriptors like Fisher Vectors or Vector of Locally Aggregated Descriptors have shown to improve the performance of many computer vision applications including retrieval. In the third part, we will discuss an unsupervised technique for compressing high dimensional vectors into high dimensional binary codes, to reduce storage complexity. In this approach, we deviate from adopting traditional hyperplane hashing functions and instead learn hyperspherical hashing functions. The proposed method overcomes the computational challenges of directly applying the spherical hashing algorithm that is intractable for compressing high dimensional vectors. A practical hierarchical model that utilizes divide and conquer techniques using the Random Select and Adjust (RSA) procedure to compress such high dimensional vectors is presented. We show that our proposed high dimensional binary codes outperform the binary codes obtained using traditional hyperplane methods for higher compression ratios. In the last part of the thesis, we propose a retrieval based solution to the Zero shot event classification problem - a setting where no training videos are available for the event. To do this, we learn a generic set of concept detectors and represent both videos and query events in the concept space. We then compute similarity between the query event and the video in the concept space and videos similar to the query event are classified as the videos belonging to the event. We show that we significantly boost the performance using concept features from other modalities.

Latent representations for traditional music analysis and generation

We create and study a generative model for Irish traditional music based on Variational Autoencoders and analyze the learned latent space trying to find musically significant correlations in the latent codes' distributions in order to perform musical analysis on data. We train two kinds of models: one trained on a dataset of Irish folk melodies, one trained on bars extrapolated from the melodies dataset, each one in five variations of increasing size. We conduct the following experiments: we inspect the latent space of tunes and bars in relation to key, time signature, and estimated harmonic function of bars; we search for links between tunes in a particular style (i.e. "reels'") and their positioning in latent space relative to other tunes; we compute distances between embedded bars in a tune to gain insight into the model's understanding of the similarity between bars. Finally, we show and evaluate generative examples. We find that the learned latent space does not explicitly encode musical information and is thus unusable for musical analysis of data, while generative results are generally good and not strictly dependent on the musical coherence of the model's internal representation.

Advancing Bag-of-visual-words Representations For Lesion Classification In Retinal Images.

Diabetic Retinopathy (DR) is a complication of diabetes that can lead to blindness if not readily discovered. Automated screening algorithms have the potential to improve identification of patients who need further medical attention. However, the identification of lesions must be accurate to be useful for clinical application. The bag-of-visual-words (BoVW) algorithm employs a maximum-margin classifier in a flexible framework that is able to detect the most common DR-related lesions such as microaneurysms, cotton-wool spots and hard exudates. BoVW allows to bypass the need for pre- and post-processing of the retinographic images, as well as the need of specific ad hoc techniques for identification of each type of lesion. An extensive evaluation of the BoVW model, using three large retinograph datasets (DR1, DR2 and Messidor) with different resolution and collected by different healthcare personnel, was performed. The results demonstrate that the BoVW classification approach can identify different lesions within an image without having to utilize different algorithms for each lesion reducing processing time and providing a more flexible diagnostic system. Our BoVW scheme is based on sparse low-level feature detection with a Speeded-Up Robust Features (SURF) local descriptor, and mid-level features based on semi-soft coding with max pooling. The best BoVW representation for retinal image classification was an area under the receiver operating characteristic curve (AUC-ROC) of 97.8% (exudates) and 93.5% (red lesions), applying a cross-dataset validation protocol. To assess the accuracy for detecting cases that require referral within one year, the sparse extraction technique associated with semi-soft coding and max pooling obtained an AUC of 94.2 ± 2.0%, outperforming current methods. Those results indicate that, for retinal image classification tasks in clinical practice, BoVW is equal and, in some instances, surpasses results obtained using dense detection (widely believed to be the best choice in many vision problems) for the low-level descriptors.

Descrições estruturais cristalinas de zeólitos

Crystalline structures of zeolites can be studied using different representations: the internal symmetry obtained by X-Ray or neutron diffraction crystallography techniques or a systematic analysis of the basic structural units which can be arranged to build the geometries of each kind of zeolite. In this work the basic concepts of three building units, SBU (Secondary Building Units), SSU (Structural SubUnits) and PBU (Periodic Building Units) are presented. The properties of the resulting crystalline structures are discussed (pores, cavities, channels), describing the influence of each one of these properties in processes of physical-chemical interest. Representative case studies of known zeolite crystalline structures are also discussed in terms of their space group classification.

Desenvolvimento de uma metodologia para reconstrução tridimensional e analise de superficies do corpo humano

Universidade Estadual de Campinas . Faculdade de Educação Física

Influence of the length of remaining root canal filling and post space preparation on the coronal leakage of Enterococcus faecalis

This study evaluated the sealing ability of different lengths of remaining root canal filling and post space preparation against coronal leakage of Enterococcus faecalis. Forty-one roots of maxillary incisors were biomechanically prepared, maintaining standardized canal diameter at the middle and coronal thirds. The roots were autoclaved and all subsequent steps were undertaken in a laminar flow chamber. The canals of 33 roots were obturated with AH Plus sealer and gutta-percha. The root canal fillings were reduced to 3 predetermined lengths (n=11): G1=6 mm, G2=4 mm and G3=2 mm. The remaining roots served as positive and negative controls. Bacterial leakage test apparatuses were fabricated with the roots attached to Eppendorf tubes keeping 2 mm of apex submerged in BHI in glass flasks. The specimens received an E. faecalis inoculum of 1 x 107 cfu/mL every 3 days and were observed for bacterial leakage daily during 60 days. Data were submitted to ANOVA, Tukey's test and Fisher's test. At 60 days, G1 (6 mm) and G2 (4 mm) presented statistically similar results (p>0.05) (54.4% of specimens with bacterial leakage) and both groups differed significantly (p<0.01) from G3 (2 mm), which presented 100% of specimens with E. faecalis leakage. It may be concluded that the shortest endodontic obturation remnant leaked considerably more than the other lengths, although none of the tested conditions avoids coronal leakage of E. faecalis.

Modelos da Química Quântica no espaço de momento: diferentes representações de um mesmo sistema

The conventional approach to simple quantum chemistry models is contrasted with that known as momentum representation, where the wavefunctions are momentum dependent. Since the physical interactions are the same, state energies should not change, and whence the energy differences correlating with the real world as spectral lines or bands. We emphasize that one representation is not more fundamental than the other, and the choice is a matter of mathematical convenience. As spatial localization is rooted in our brains, to think in terms of the momentum present us a great mental challenge that can lead to complementary perspectives of a model.

The bubbles or the boiling pot?: an ecosystemic approach to culture, environment and quality of life

For the diagnosis and prognosis of the problems of quality of life, a multidisciplinary ecosystemic approach encompasses four dimensions of being-in-the-world, as donors and recipients: intimate, interactive, social and biophysical. Social, cultural and environmental vulnerabilities are understood and dealt with, in different circumstances of space and time, as the conjugated effect of all dimensions of being-in-the-world, as they induce the events (deficits and assets), cope with consequences (desired or undesired) and contribute for change. Instead of fragmented and reduced representations of reality, diagnosis and prognosis of cultural, educational, environmental and health problems considers the connections (assets) and ruptures (deficits) between the different dimensions, providing a planning model to develop and evaluate research, teaching programmes, public policies and field projects. The methodology is participatory, experiential and reflexive; heuristic-hermeneutic processes unveil cultural and epistemic paradigms that orient subject-object relationships; giving people the opportunity to reflect on their own realities, engage in new experiences and find new ways to live better in a better world. The proposal is a creative model for thought and practice, providing many opportunities for discussion, debate and development of holistic projects integrating different scientific domains (social sciences, psychology, education, philosophy, etc.)

A questão social na França contemporânea: uma experiência de pesquisa

A França passou por conflitos urbanos em seus bairros populares. Incidentes com a polícia marcados pela morte de um morador, geralmente jovem ou adolescente, são seguidos por manifestações violentas: quebra-quebras, incêndios de carros e/ou estabelecimentos públicos e privados, enfrentamentos com a polícia. Nos conflitos da "banlieue" francesa podem ser percebidos elementos comuns a eventos ingleses e/ou americanos de outras décadas.OBJETIVO: consiste em identificar a percepção e as representações de moradores sobre suas condições de vida e saúde na França.MÉTODO: pesquisa qualitativa realizada em Les Aubiers, bairro popular de Bordeaux, nos anos de 2002 e 2005. A população de moradores desse bairro e é composta principalmente por imigrantes ou fanceses de origem estrangeira, provenientes de diferentes locais: África do Norte, África Sub-Sahariana, Turquia, Portugal , Ilha de Reunião. Os indicadores sociais e econômicos apresentam níveis significativamente mais baixos do que os de outros bairros de Bordeaux, acompanhando as tendências encontradas nas outras "banlieues". Entre esses indicadores, a taxa de desemprego, principalmente entre jovens, é alarmante (40 por cento).RESULTADOS E DISCUSSÃO: as políticas sociais deveriam ser mais voltadas para os indivíduos do que para o território, pois este só poderá ser transformado pelos primeiros e não o contrário. Um bairro e uma população estigmatizados, estão contemplados por uma ampla política social. Ainda que passível de crítica, não se pode ignorar a existência de esforços sistematicos de integração, nem a existência de políticas sociais. Se elas são inadequadas, trata-se de ampliar o espaço de discusão para sua reforma. Trata-se de saber qual a disposição das sociedades contemporâneas, como um todo, para lutar contra as inequidades e segregações

Quasinormal modes of black holes in anti-de Sitter space: A numerical study of the eikonal limit

Using series solutions and time-domain evolutions, we probe the eikonal limit of the gravitational and scalar-field quasinormal modes of large black holes and black branes in anti-de Sitter backgrounds. These results are particularly relevant for the AdS/CFT correspondence, since the eikonal regime is characterized by the existence of long-lived modes which (presumably) dominate the decay time scale of the perturbations. We confirm all the main qualitative features of these slowly damped modes as predicted by Festuccia and Liu [G. Festuccia and H. Liu, arXiv:0811.1033.] for the scalar-field (tensor-type gravitational) fluctuations. However, quantitatively we find dimensional-dependent correction factors. We also investigate the dependence of the quasinormal mode frequencies on the horizon radius of the black hole (brane) and the angular momentum (wave number) of vector- and scalar-type gravitational perturbations.

TRAVELLING THROUGH PITCH SPACE SPEEDS UP MUSICAL TIME

SEVERAL MODELS OF TIME ESTIMATION HAVE BEEN developed in psychology; a few have been applied to music. In the present study, we assess the influence of the distances travelled through pitch space on retrospective time estimation. Participants listened to an isochronous chord sequence of 20-s duration. They were unexpectedly asked to reproduce the time interval of the sequence. The harmonic structure of the stimulus was manipulated so that the sequence either remained in the same key (CC) or travelled through a closely related key (CFC) or distant key (CGbC). Estimated times were shortened when the sequence modulated to a very distant key. This finding is discussed in light of Lerdahl's Tonal Pitch Space Theory (2001), Firmino and Bueno's Expected Development Fraction Model (in press), and models of time estimation.

The role of heritage in revalorization politicos of urban space

Certain areas of the city of Sao Paulo, as many others around the world, including in Lisbon, Barcelona and Buenos Aires, have been going through a process of requalification, in special the ones commonly known as old and/or traditional city. Regarding Sao Paulo, some exceptional actions have been taken downtown with investments in rehabilitation/requalification of areas that concentrated the historical, urbanistic and cultural heritages, such as Praca da S and its cathedral, as well as the revaluation/rehabilitation projects of other squares like Praca da Republica, other areas as the previously called Cracolandia (due to high consumption/deal of crack), known today as Nova Luz, besides propositions to reevaluate areas already modified, such as Vale do Anhangabau. In all propositions to modify sites, it is firstly underlined its deterioration, litter and the presence of low-income populations (passer-bys, street vendors or residents), generally stigmatized as ""potential suspects"", emphasizing danger and lack of security in those places. This belief, which has become consensual, results in that: public as well as private companies promote the rehabilitation of the areas basing their reasoning in the necessity of adding value to the existing urban heritage, although, as it will be discussed in this paper, part of this heritage might be destroyed in this very process, under the allegation that upon completion, the action would allow the social, cultural and economical revaluation/requalification of the area. This paper is intended to provide a contribution to this discussion.

Transiting exoplanets from the CoRoT space mission XVI. CoRoT-14b: an unusually dense very hot Jupiter

In this paper, the CoRoT Exoplanet Science Team announces its 14th discovery. Herein, we discuss the observations and analyses that allowed us to derive the parameters of this system: a hot Jupiter with a mass of 7.6 +/- 0.6 Jupiter masses orbiting a solar-type star (F9V) with a period of only 1.5 d, less than 5 stellar radii from its parent star. It is unusual for such a massive planet to have such a small orbit: only one other known higher mass exoplanet orbits with a shorter period.

Transiting exoplanets from the CoRoT space mission XVIII. CoRoT-18b: a massive hot Jupiter on a prograde, nearly aligned orbit

We report the detection of CoRoT-18b, a massive hot Jupiter transiting in front of its host star with a period of 1.9000693 +/- 0.0000028 days. This planet was discovered thanks to photometric data secured with the CoRoT satellite combined with spectroscopic and photometric ground-based follow-up observations. The planet has a mass M(p) = 3.47 +/- 0.38 M(Jup), a radius R(p) = 1.31 +/- 0.18 R(Jup), and a density rho(p) = 2.2 +/- 0.8 g cm(-3). It orbits a G9V star with a mass M(*) = 0.95 +/- 0.15 M(circle dot), a radius R(*) = 1.00 +/- 0.13 R(circle dot), and a rotation period P(rot) = 5.4 +/- 0.4 days. The age of the system remains uncertain, with stellar evolution models pointing either to a few tens Ma or several Ga, while gyrochronology and lithium abundance point towards ages of a few hundred Ma. This mismatch potentially points to a problem in our understanding of the evolution of young stars, with possibly significant implications for stellar physics and the interpretation of inferred sizes of exoplanets around young stars. We detected the RossiterMcLaughlin anomaly in the CoRoT-18 system thanks to the spectroscopic observation of a transit. We measured the obliquity psi = 20 degrees +/- 20 degrees +/- (sky-projected value lambda = -10 degrees +/- 20 degrees), indicating that the planet orbits in the same way as the star is rotating and that this prograde orbit is nearly aligned with the stellar equator.

Transiting exoplanets from the CoRoT space mission XV. CoRoT-15b: a brown-dwarf transiting companion

We report the discovery by the CoRoT space mission of a transiting brown dwarf orbiting a F7V star with an orbital period of 3.06 days. CoRoT-15b has a radius of 1.12(-0.15)(+0.30) R(Jup) and a mass of 63.3 +/- 4.1 M(Jup), and is thus the second transiting companion lying in the theoretical mass domain of brown dwarfs. CoRoT-15b is either very young or inflated compared to standard evolution models, a situation similar to that of M-dwarf stars orbiting close to solar-type stars. Spectroscopic constraints and an analysis of the lightcurve imply a spin period in the range 2.9-3.1 days for the central star, which is compatible with a double-synchronisation of the system.