Datasets, features, learning, and models in visual recognition

Visual recognition is a fundamental research topic in computer vision. This dissertation explores datasets, features, learning, and models used for visual recognition. In order to train visual models and evaluate different recognition algorithms, this dissertation develops an approach to collect object image datasets on web pages using an analysis of text around the image and of image appearance. This method exploits established online knowledge resources (Wikipedia pages for text; Flickr and Caltech data sets for images). The resources provide rich text and object appearance information. This dissertation describes results on two datasets. The first is Berg’s collection of 10 animal categories; on this dataset, we significantly outperform previous approaches. On an additional set of 5 categories, experimental results show the effectiveness of the method. Images are represented as features for visual recognition. This dissertation introduces a text-based image feature and demonstrates that it consistently improves performance on hard object classification problems. The feature is built using an auxiliary dataset of images annotated with tags, downloaded from the Internet. Image tags are noisy. The method obtains the text features of an unannotated image from the tags of its k-nearest neighbors in this auxiliary collection. A visual classifier presented with an object viewed under novel circumstances (say, a new viewing direction) must rely on its visual examples. This text feature may not change, because the auxiliary dataset likely contains a similar picture. While the tags associated with images are noisy, they are more stable when appearance changes. The performance of this feature is tested using PASCAL VOC 2006 and 2007 datasets. This feature performs well; it consistently improves the performance of visual object classifiers, and is particularly effective when the training dataset is small. With more and more collected training data, computational cost becomes a bottleneck, especially when training sophisticated classifiers such as kernelized SVM. This dissertation proposes a fast training algorithm called Stochastic Intersection Kernel Machine (SIKMA). This proposed training method will be useful for many vision problems, as it can produce a kernel classifier that is more accurate than a linear classifier, and can be trained on tens of thousands of examples in two minutes. It processes training examples one by one in a sequence, so memory cost is no longer the bottleneck to process large scale datasets. This dissertation applies this approach to train classifiers of Flickr groups with many group training examples. The resulting Flickr group prediction scores can be used to measure image similarity between two images. Experimental results on the Corel dataset and a PASCAL VOC dataset show the learned Flickr features perform better on image matching, retrieval, and classification than conventional visual features. Visual models are usually trained to best separate positive and negative training examples. However, when recognizing a large number of object categories, there may not be enough training examples for most objects, due to the intrinsic long-tailed distribution of objects in the real world. This dissertation proposes an approach to use comparative object similarity. The key insight is that, given a set of object categories which are similar and a set of categories which are dissimilar, a good object model should respond more strongly to examples from similar categories than to examples from dissimilar categories. This dissertation develops a regularized kernel machine algorithm to use this category dependent similarity regularization. Experiments on hundreds of categories show that our method can make significant improvement for categories with few or even no positive examples.

WebQuest: Incidencia en el aprendizaje significativo para la educación superior

El presente artículo explora el uso de la WebQuest como metodología del Aprendizaje Basado en Proyectos Colaborativos en la formación de estudiantes de educación superior de la Escuela de Ingeniería Mecánica de la Escuela Superior Politécnica de Chimborazo (ESPOCH). Con esta finalidad se diseñó la WebQuest para ser aplicada con estudiantes entre 18 y 22 años en la asignatura de Programación. La investigación fue de tipo cognitivo-constructivista, descriptiva, de campo, transversal, aplicada, de ciencias básicas y bajo el paradigma científico, enmarcada bajo políticas y lineamientos del Plan Nacional del Buen Vivir 2013-2017. Mediante la utilización de herramientas colaborativas Google Drive y Google Sites, el test CHAEA para estilos de aprendizaje, herramientas ofimáticas y software especializado determinando así su incidencia en el aprendizaje significativo de los 54 estudiantes tomados como muestra. Se aplicaron dos técnicas estadísticas: t-Student para comprobar el rendimiento académico, comportamiento y estilos de aprendizaje. Con 5% de error y un punto crítico de 2,055 a dos colas, se obtuvieron valores de: 4,382, 2,434 y 3,701 respectivamente. La técnica Chi-cuadrado se aplicó para comprobar las habilidades del siglo XXI, con un punto crítico de 40,1, se generó el valor calculado de 119,84, demostrando que la metodología WebQuest favorece la generación de aprendizajes significativos dentro y fuera del aula de clase.

Characterization of 27 microsatellite loci in the European flat oyster Ostrea edulis

The flat oyster Ostrea edulis is native to Europe and populations have been severely depleted by the parasite Bonamia ostreae since the 1980s. Additional genetic markers are required to improve population genetics study and linkage map development for selection for B. ostrea-resistance in this species. Here, we characterized 27 novel microsatellite loci for O. edulis. Number of alleles per locus ranged from 6 to 25 and observed heterozygosity between 0.375 and 1. Null alleles were suggested at a few loci but most loci were in Hardy-Weinberg agreement enabling their reliable use in further population and mapping genetics approaches.

BALD EAGLES (HALIAEETUS LEUCOCEPHALUS) AS INDICATORS OF GREAT LAKES ECOSYSTEM HEALTH

Environmental indicators have been proposed as a means to assess ecological integrity, monitoring both chemical and biological stressors. In this study, we used nestling bald eagles as indicators to quantify direct or indirect tertiary-level contaminant exposure. The spatial and temporal trends of polychlorinated biphenyl (PCB) congeners were evaluated in nestling plasma from 1999–2014. Two hexa-chlorinated congeners, PCB-138 and 153, were detected with the highest frequency and greatest concentrations throughout Michigan. Less-chlorinated congeners such as PCB-52 and 66 however, comprised a greater percentage of total PCB concentrations in nestlings proximate to urbanized areas, such as along the shorelines of Lake Erie. Toxic equivalents were greatest in the samples collected from nestlings located on Lake Erie, followed by the other Great Lakes spatial regions. Nestling plasma samples were also used to measure concentrations of the most heavily-used group of flame retardants, brominated diphenyl ethers (BDEs), and three groups of alternative flame retardants, non-BDE Brominated Flame Retardants (NBFRS), Dechloranes, and organophosphate esters (OPs). BDE-47, 99 and 100 contributed the greatest to total BDE concentrations. Concentrations of structurally similar NBFRs found in this study and recent atmospheric studies indicate that they are largely used as replacements to previously used BDE mixtures. A variety of Dechloranes, or derivatives of Mirex and Dechlorane Plus, were measured. Although, measured at lesser concentrations, environmental behavior of these compounds may be similar to mirex and warrant future research in aquatic species. Concentrations of OPs in nestling plasma were two to three orders of magnitude greater than all other groups of flame retardants. In addition to chemical indicators, bald eagles have also been proposed as indicators to identify ecological stressors using population measures that are tied to the fitness of individuals and populations. Using mortality as a population vitality rate, vehicle collisions were found to be the main source of mortality with a greater incidence for females during white-tailed deer (Odocoileus virginianus) hunting months and spring snow-melt. Lead poisoning was the second greatest source of mortality, with sources likely due to unretrieved hunter-killed, white-tailed deer carcasses, and possibly exacerbated by density-dependent effects due to the growing population in Michigan.

Analisis bromatológico inter-laboratorio de harina de centeno a ser utilizado como patrón secundario de análisis

El objetivo de esta investigación fue realizar un estudio inter-laboratorio bromatológico a partir de un patrón secundario de harina de centeno. Los laboratorios participantes de este estudio comparativo fueron el Laboratorio de Análisis Bromatológico para atención al público de la Universidad de Cuenca y el Laboratorio de Alimentos y Nutrición (VLIR-IUC) del Departamento de Biociencias. Los parámetros del análisis fueron: humedad, materia seca, cenizas, proteína bruta total, grasa total y carbohidratos totales. También el contenido de sal (cloruros) se analizó en uno de los laboratorios. Para los análisis de los diferentes parámetros se utilizaron las metodologías establecidas en cada laboratorio.Con los datos generados se realizaron gráficas de controlLevey-Jennings para cada parámetro y laboratoriopara el posterior control de análisis utilizando el patrón secundario dentro de los dos laboratorios participantes. Los resultados fueron evaluados estadísticamente mediante pruebas T de Student de una cola utilizando un nivel de significancia del 5%. Además se determinó la precisión intra- e inter-día siguiendo el método ANOVA y se expresó como porcentaje de coeficiente de variación (% CV). Todos los análisis fueron realizados en los programas Microsoft Excel 2013 y STATA 10.0. Para ambos laboratorios, los coeficientes de variación inter- e intra-día no sobrepasaron el 15%, que es lo recomendado para análisis proximal. Por otro lado, se encontraron diferencias significativas en los análisis de grasas, humedad-materia seca y carbohidratos, que pueden atribuirse a las diferencias metodológicas, temperatura y equipos de análisis, y en menor grado a los analistas y al azar.

SVR-GARCH com misturas de kernels gaussianos

Dissertação (mestrado)—Universidade de Brasília, Departamento de Administração, Programa de Pós-graduação em Administração, 2016.

Functional and structural dissection of the tape measure protein of lactococcal phage TP901-1

The tail tape measure protein (TMP) of tailed bacteriophages (also called phages) dictates the tail length and facilitates DNA transit to the cell cytoplasm during infection. Here, a thorough mutational analysis of the TMP from lactococcal phage TP901-1 (TMPTP901-1) was undertaken. We generated 56 mutants aimed at defining TMPTP901-1 domains that are essential for tail assembly and successful infection. Through analysis of the derived mutants, we determined that TP901-1 infectivity requires the N-terminal 154 aa residues, the C-terminal 60 residues and the first predicted hydrophobic region of TMPTP901-1 as a minimum. Furthermore, the role of TMPTP901-1 in tail length determination was visualized by electron microscopic imaging of TMP-deletion mutants. The inverse linear correlation between the extent of TMPTP901-1-encoding gene deletions and tail length of the corresponding virion provides an estimate of TMPTP901-1 regions interacting with the connector or involved in initiator complex formation. This study represents the most thorough characterisation of a TMP from a Gram-positive host-infecting phage and provides essential advances to understanding its role in virion assembly, morphology and infection.

Measuring Latency Variation in the Internet

We analyse two complementary datasets to quantify the latency variation experienced by internet end-users: (i) a large-scale active measurement dataset (from the Measurement Lab Network Diagnostic Tool) which shed light on long-term trends and regional differences; and (ii) passive measurement data from an access aggregation link which is used to analyse the edge links closest to the user. The analysis shows that variation in latency is both common and of significant magnitude, with two thirds of samples exceeding 100\,ms of variation. The variation is seen within single connections as well as between connections to the same client. The distribution of experienced latency variation is heavy-tailed, with the most affected clients seeing an order of magnitude larger variation than the least affected. In addition, there are large differences between regions, both within and between continents. Despite consistent improvements in throughput, most regions show no reduction in latency variation over time, and in one region it even increases. We examine load-induced queueing latency as a possible cause for the variation in latency and find that both datasets readily exhibit symptoms of queueing latency correlated with network load. Additionally, when this queueing latency does occur, it is of significant magnitude, more than 200\,ms in the median. This indicates that load-induced queueing contributes significantly to the overall latency variation.

2016 Flanders Virus Cases

Flanders virus was discovered in 1961 in the town of Flanders on Long Island, New York. The virus is in the virus family Rhabdoviridae, and it is widely distributed in Canada, the United States, and Mexico. Flanders virus does not cause disease. The virus is frequently found in birds, such as Red-winged blackbirds, House sparrows, and starlings. It is also found in bird-feeding mosquitoes, such as the black-tailed mosquito (Culiseta melanura) and the northern and southern house mosquitoes (Culex pipiens and Culex quinquefasciatus, respectively). Presence of the virus in an area serves as a sentinel or warning for West Nile virus and as a trigger for public health control and prevention interventions targeting West Nile virus.

Sequence analysis of bitter taste receptor gene repertoires in different ruminant species

Bitter taste has been extensively studied in mammalian species and is associated with sensitivity to toxins and with food choices that avoid dangerous substances in the diet. At the molecular level, bitter compounds are sensed by bitter taste receptor proteins (T2R) present at the surface of taste receptor cells in the gustatory papillae. Our work aims at exploring the phylogenetic relationships of T2R gene sequences within different ruminant species. To accomplish this goal, we gathered a collection of ruminant species with different feeding behaviors and for which no genome data is available: American bison, chamois, elk, European bison, fallow deer, goat, moose, mouflon, muskox, red deer, reindeer and white tailed deer. The herbivores chosen for this study belong to different taxonomic families and habitats, and hence, exhibit distinct foraging behaviors and diet preferences. We describe the first partial repertoires of T2R gene sequences for these species obtained by direct sequencing. We then consider the homology and evolutionary history of these receptors within this ruminant group, and whether it relates to feeding type classification, using MEGA software. Our results suggest that phylogenetic proximity of T2R genes corresponds more to the traditional taxonomic groups of the species rather than reflecting a categorization by feeding strategy.

USO DE SISTEMAS DE INFORMACION GEOGRAFICA PARA ANALIZAR IMPACTOS DE ACTIVIDADES HUMANAS SOBRE EL HABITAT POTENCIAL DEL VENADO COLA BLANCA (odocoileus virginianus), GUANACASTE, COSTA RICA.

El presente trabajo se realizó en el distrito central del cantón de Bagaces, Guanacaste, Costa Rica y tiene como propósito evaluar el impacto de la II fase del proyecto de Riego Arenal-Tempisque en el hábitat potencial del venado cola blanca (odocoileus virginianus). Los resultados del estudio indican que la II etapa del Proyecto de Riego impactará 4.633 hectáreas del hábitat clasificado como alto para el venado. SUMMARY The aim of this paper is to discuss and illustrate the usefulness of the GIS in the evaluation of the impact of the II Phase of the Arenal-Tempisque irrigation project on the habitat of white-tailed deer. The study area is the central canton of Bagaces in Guanacaste, Costa Rica. Once fully developed, the irrigation project will impact 4633 hectares of prime deer habitat.