In this thesis, we present ContextNet, a novel general object detection framework for incorporating context cues into a detection pipeline. Current deep learning methods for object detection exploit state-of-the-art image recognition networks for classifying the given region-of-interest (ROI) to predefined classes and regressing a bounding-box around it without using any information about the corresponding scene. ContextNet is based on an intuitive idea of having cues about the general scene (e.g., kitchen and library), and changes the priors about presence/absence of some object classes. We provide a general means for integrating this notion in the decision process about the given ROI by using a pretrained network on the scene recognition datasets in parallel to a pretrained network for extracting object-level features for the corresponding ROI. Using comprehensive experiments on the PASCAL VOC 2007, we demonstrate the effectiveness of our design choices, the resulting system outperforms the baseline in most object classes, and reaches 57.5 mAP (mean Average Precision) on the PASCAL VOC 2007 test set in comparison with 55.6 mAP for the baseline. / MS / The object detection problem is to find objects of interest in a given image and draw boxes around them with object labels. With the emergence of deep learning in recent years, current object detection methods use deep learning technologies. The detection process is solely based on features which are extracted from several thousand regions in the given image. We propose a novel framework for incorporating scene information in the detection process. For example, if we know the image is taken from a kitchen, the probability of seeing a cow or an airplane decreases and observation probability of plates and persons increases. Our new detection network uses this intuition to improve the detection accuracy. Using extensive experiments, we show the proposed methods outperform the baseline for almost all object types.

Call number: LD2668 .T4 1985 S36 / Master of Science

Master of Science / Department of Biology / Kimberly A. With / Organisms exist within complex landscapes, and landscape features may influence multiple aspects of a species’ distribution within the landscape, including patch occupancy, abundance within patches, and population genetic diversity at a local or regional scale. We took two approaches to identify the relative importance of landscape context for populations of the Eastern Collared Lizard (Crotaphytus collaris collaris) in the northern Flint Hills of Kansas. First, we conducted surveys at limestone outcrops in experimental watersheds managed under different burning and grazing practices. Habitat occupancy and lizard abundance were estimated by constructing models that incorporated aspects of the environment at multiple scales. Both abundance and occupancy were higher on rock ledges that had more crevices, greater complexity in vegetation, covered a larger area, and were more prominent than available habitat. Abundance and occupancy were also higher in watersheds that were burned frequently (1–2 year intervals), but grazing only had a significant effect in less frequently burned (four–year burn interval) watersheds. Our second approach was to measure genetic diversity and population genetic differentiation and relate these measures to differences in landscape context. We sampled collared lizard DNA at four locations (sample sites < 45 km apart) and analyzed trends in 10 polymorphic microsatellite loci. We found evidence for low genetic variability and moderate population differentiation among our sample sites relative to estimates reported in the literature at the core of the species’ range. Differences in migration rates and ancestry among sampling locations also appear to correspond to differences in landscape resistance based on land cover and rock availability. Thus, it appears that habitat management may influence the suitability of habitat patches at the local scale, and that differences in land cover and rock availability may influence the connectivity of populations at the landscape scale.

An ambient odour of anise was used in a context-dependent memory study with three different memory tasks targeting both declarative and non-declarative memory functions. Declarative memory was assessed by means of two episodic memory tests; recall of a prose text and a complex figure. Priming was used to assess the non-declarative memory with word fragment completion. Memory was tested immediately and after 48 hours. The results showed a significant main effect of context (odour or not) for all three tests in favour of the olfactory context. In addition, a significant main effect of time was observed for all three tests and a significant interaction effect between context and time for the priming test were observed. This interaction showed that the priming effect was equal in size across both conditions at immediate testing, although when the odour was reinstated at the delayed test the results showed larger priming relative the control condition.

Sammanfattning Denna studie avser att undersöka hur socialsekreterare, boendepersonal och ensamkommande barn talar om integration samt vilka skillnader och likheter som konstrueras mellan grupperna. Syftet är även att studera vilka problem som uppstår i och med att integrationen konstrueras på olika sätt. Detta görs genom kvalitativa intervjuer med tre ensamkommande ungdomar, tre socialsekreterare och fyra boendepersonal. Materialet bearbetas utifrån ett diskursanalytiskt perspektiv och belyser konstruktioner som vi ansett haft en betydelse för våra frågeställningar: Diskursanalyshar även varit vår teoretiska utgångspunkt tillsammans med José Alberto Diazactor context model. Studien visar att det inte finns en enhetlig konstruktion av integration, vilket gör att arbetet och förväntningarna kring integration ser olika ut. Genom respondenternas konstruktioner belyser vi olika aspekter som de menar har betydelse för integrationsprocessen. Deras konstruktioner ger även en bild på förväntningar och behov som respondenterna menar måste tillgodoses för ett gynnsammare arbete med ensamkommande barn och dess integration.

Vehicular Ad-hoc Network (VANET) is a special application of the Mobile Ad-hoc Network (MANET) for managing road traffic and substantially contributes to the development of Intelligent Transportation Systems (ITS). VANET was introduced as a standard for data communication between moving vehicles with and without fixed infrastructure. It aims to support drivers by improving safety and driving comfort as a step towards constructing a safer, cleaner and a more intelligent environment. Nowadays, vehicles are manufactured equipped with a number of sensors and devices called On Board Units (OBU) assisting the vehicle to sense the surrounding environment and then process the context information to effectively manage communication with the surrounding vehicles and the associated infrastructure. A number of challenges have emerged in VANET that have encouraged researchers to investigate this concept further. Many of the recent studies have applied different technologies for intelligent parking management. However, despite all the technological advances, researchers are no closer to developing a system that enables drivers to easily locate and reserve a parking space. Limited resources such as energy, storage space, availability and reliability are factors which could have contributed to the lack success and progress in this area. The task then is to close these gaps and present a novel solution for parking. This research intends to address this need by developing a novel architecture for locating and reserving a parking space that best matches the driver's preferences and vehicle profile without distracting the driver. The simple and easy-to-use mechanism focuses on the domain of an intelligent parking system that exploits the concept of InfoStation (IS) and context-aware system creating a single framework to locate and reserve a parking space. A three tier network topology comprising of vehicles, IS and the InfoStation Centre (ISC) has been proposed as the foundation of the on-street parking system architecture. The thesis attempts to develop the architecture of a parking management solution as a comfort-enhancing application that offers to reduce congestion related stress and improve the driver experience by reducing the time it takes to identify and utilise a parking space that is available.

Syftet med denna studie var att undersöka hur lycka upplevs av förskolebarn.  Individuella intervjuer gjordes med 26 förskolebarn (73% pojkar) mellan 4 och 5 år, med hjälp av en semistrukturerad intervjuguide. Tidigare forskning har visat på att lycka har många delar så som kärlek, välbefinnande, tillfredställelse, god hälsa och att lycka skulle kunna vara något större så som meningen med livet. Materialet analyserades med stöd av ICF-CY och resultatet visade att barn upplevde lycka när de kände sig delaktiga i ett sammanhang. Omgivningsfaktorer, så som att barnen fick något av materiell karaktär samt familjekontexten har också betydelse om barn känner sig lyckliga. Barn ville känna sig delaktiga, ha en stabil omgivning och kärleksfulla nära relationer. Denna studie bidrar till att belysa barns egen uppfattning av just begreppet lycka samt ger kunskap som bidrar till att kunna upprätthålla barns chans att få känna lycka. Vidare forskning skulle vara att undersöka om olika personlighets typer  har olika uppfattning om lycka.

The purpose of the present study was to determine whether computer-based cognitive bias modification (CBM) procedures could alter appraisal style toward viewing events from a big picture perspective and thereby influence emotional reactivity. Big picture appraisal entails viewing difficult situations and one's reactions to them in terms of a larger context. Appraisal training was implicit in that participants completed a series of vignettes, framed as a reading comprehension task, which trained either a big picture perspective or a personal/evaluative focus. When subsequently confronted with novel vignettes, participants produced interpretations that were consistent with assigned training condition. In addition, participants trained in big picture as compared to personal/evaluative appraisal subsequently demonstrated less emotional reactivity to a stressful task. / text

I examine the role of context as a rhetorical trope. As a rhetorical trope, context tends to fix complex practices in single places, which allows for the celebration of the authentic or original. Further, it privileges production while masking complex practices of circulation and consumption while simultaneously constraining seemingly infinite possibilities into finite frames that then become static and naturalized. These practices need to be examined in order to understand how power is being enacted via the trope of context for the purposes of control and limitation. I argue throughout that these power dynamics need to be addressed--that the ethics of context need to consider who or what is empowered, who or what is disempowered, and decide whether such a situational power dynamic is acceptable or should be changed.I move through the dissertation by first presenting the metaphors of context--maps, frames, and landscapes--discussing the ways in which each of these metaphors control and limit context and therefore control and limit the text. I then analyze the textual and rhetorical context traditions to illuminate the ways in which these two prevalent traditions assume a static and constant original context to which a text belongs. The constant appeal to an origin, I argue, invests a text or artifact with historical aura, which is often used to obscure and limit other critical engagements with a text thus controlling a text's or artifact's possible meanings and transformative power. Following this exploration, I turn my attention to contexts as consumable commodities. I argue that contexts as rhetorical tropes are divorced from the dialectical process of meaning making from a text and can therefore exist as its own entity. As such, contexts can be marketed to and consumed by people. An ethics of context, I conclude, would challenge the god term that context has become in order to expose the power and ideological control that is exerted via a deployment of rhetorical contexts. Such an ethics would address, again, the dialectical formation of texts and contexts--texts define contexts; contexts define texts; they are inseparable.

Previous research has investigated intentional retrieval of contextual information and contextual influences on object identification and word recognition, yet few studies have systematically investigated context effects in episodic memory for objects. To address this issue, unique objects on a white background or embedded in a visually rich context were presented to participants. At test, the object was presented either in the original or a different context. Chapter 2 demonstrated that a context shift decrement (CSD)--decreased recognition performance when context is changed between encoding and retrieval--was observed. In four studies with young adults, the CSD was not attenuated by encoding or retrieval manipulations. Chapter 3 revealed that the CSD was resistant to aging and neuropsychological status. Importantly, older adults classified as high MTL performed better on the recognition task than those classified as low MTL, and as well as young adults, supporting the successful aging hypothesis. Chapter 4 focused on elucidating the neural correlates of the CSD using functional Magnetic Resonance Imaging. Right PHG activation during encoding was associated with subsequent recognition of objects in the context change condition. This same region was activated during recognition, suggesting it may automatically reinstate visual contextual information. Overall, the CSD is attributed to the automatic and obligatory binding of object and context information in episodic memory that results in an integrated representation, mediated by the hippocampal complex.