Thesis (Ph.D.)--University of Florida, 2005. / Title from title page of source document. Document formatted into pages; contains 101 pages. Includes vita. Includes bibliographical references.

Thesis (M.Eng.)--Memorial University of Newfoundland, 2001. / Bibliography: leaves 108-112.

Thesis (Ph. D.)--University of Washington, 2001. / Vita. Includes bibliographical references (p. 169-184).

Thesis (M. Phil.)--Hong Kong University of Science and Technology, 2003. / Includes bibliographical references (leaves 66-68). Also available in electronic version. Access restricted to campus users.

Includes bibliographical references (p. 65-67).

Thesis (M.S.)--Kutztown University of Pennsylvania, 1995. / Source: Masters Abstracts International, Volume: 45-06, page: 2711. Typescript. Abstract appears on leaves 2-3. Includes bibliographical references (leaves 65-67).

The current trend in visual recognition research is to place a strict division between the supervised and unsupervised learning paradigms, which is problematic for two main reasons. On the one hand, supervised methods require training data for each and every category that the system learns; training data may not always be available and is expensive to obtain. On the other hand, unsupervised methods must determine the optimal visual cues and distance metrics that distinguish one category from another to group images into semantically meaningful categories; however, for unlabeled data, these are unknown a priori. I propose a visual category discovery framework that transcends the two paradigms and learns accurate models with few labeled exemplars. The main insight is to automatically focus on the prevalent objects in images and videos, and learn models from them for category grouping, segmentation, and summarization. To implement this idea, I first present a context-aware category discovery framework that discovers novel categories by leveraging context from previously learned categories. I devise a novel object-graph descriptor to model the interaction between a set of known categories and the unknown to-be-discovered categories, and group regions that have similar appearance and similar object-graphs. I then present a collective segmentation framework that simultaneously discovers the segmentations and groupings of objects by leveraging the shared patterns in the unlabeled image collection. It discovers an ensemble of representative instances for each unknown category, and builds top-down models from them to refine the segmentation of the remaining instances. Finally, building on these techniques, I show how to produce compact visual summaries for first-person egocentric videos that focus on the important people and objects. The system leverages novel egocentric and high-level saliency features to predict important regions in the video, and produces a concise visual summary that is driven by those regions. I compare against existing state-of-the-art methods for category discovery and segmentation on several challenging benchmark datasets. I demonstrate that we can discover visual concepts more accurately by focusing on the prevalent objects in images and videos, and show clear advantages of departing from the status quo division between the supervised and unsupervised learning paradigms. The main impact of my thesis is that it lays the groundwork for building large-scale visual discovery systems that can automatically discover visual concepts with minimal human supervision. / text

Bandwidth-intensive video streaming applications occupy an overwhelming fraction of bandwidth-limited wireless network traffic. Compressed video data are highly structured and the psycho-visual perception of distortions and losses closely depends on that structure. This dissertation exploits the inherent video data structure to develop perceptually-optimized transmission paradigms at different protocol layers that improve video quality of experience, introduce error resilience, and enable supporting more video users. First, we consider the problem of network-wide perceptual quality optimization whereby different video users with (possibly different) real-time delay constraints are sharing wireless channel resources. Due to the inherently stochastic nature of wireless fading channels, we provide statistical delay guarantees using the theory of effective capacity. We derive the resource allocation policy that maximizes the sum video quality and show that the optimal operating point per user is such that the rate-distortion slope is the inverse of the supported video source rate per unit bandwidth, termed source spectral efficiency. We further propose a scheduling policy that maximizes the number of scheduled users that meet their QoS requirement. Next, we develop user-level perceptual quality optimization techniques for non-scalable video streams. For non-scalable videos, we estimate packet loss visibility through a generalized linear model and use for prioritized packet delivery. We solve the problem of mapping video packets to MIMO subchannels and adapting per-stream rates to maximize the total perceptual value of successfully delivered packets per unit time. We show that the solution enables jointly reaping gains in terms of improved video quality and lower latency. Optimized packet-stream mapping enables transmission of more relevant packets over more reliable streams while unequal modulation opportunistically increases the transmission rate on the stronger streams to enable low latency delivery of high priority packets. Finally, we develop user-level perceptual quality optimization techniques for scalable video streams. We propose online learning of the mapping between packet losses and quality degradation using nonparametric regression. This quality-loss mapping is subsequently used to provide unequal error protection for different video layers with perceptual quality guarantees. Channel-aware scalable codec adaptation and buffer management policies simultaneously ensure continuous high-quality playback. Across the various contributions, analytic results as well as video transmission simulations demonstrate the value of perceptual optimization in improving video quality and capacity. / text

The ability to make purchases in community settings is highly advantageous as it allows individuals freedom of choice and the ability to function within their own community. Independence and autonomy is especially important for individuals with intellectual disabilities (ID’s), however prerequisite knowledge of money concepts required for making cash purchases may be too complex for individuals with cognitive challenges. The use of EFTPOS cards to make purchases is a comparatively easy process with limited prerequisite skills required therefore, is an ideal starting point for teaching purchasing skills to individuals with cognitive challenges. Video modelling (VM) and video self-modelling (VSM) procedures have shown to be effective and efficient instructional techniques for teaching various skills to individuals with ID’s however, research on the effectiveness and efficiency of these procedures with individuals with Down syndrome (DS) or with EFTPOS purchases is minimal. This study aimed to examine the effectiveness of VM and VSM interventions in teaching independent EFTPOS purchasing skills to 6 young adults with DS using a non-concurrent within-participant design. The results indicates that both VM and VSM interventions were effective and efficient as all 6 participants exhibited increases in task acquisition with the introduction of the intervention, and 5 of the 6 were able to consistently use their EFTPOS cards to purchase chosen items throughout intervention and follow-up generalisation probes (2 weeks postintervention). Therefore, this study suggests both VM and VSM may be equally effective for teaching young adults with DS EFTPOS purchasing skills in community stores.

Online streaming service sites such as YouTube and Hulu are popular these days. The fact that these websites utilize online video advertising as a revenue model led to a natural curiosity: Is online video advertising effective? Although this report does not give a conclusive answer due to lack of available data, it still covers enough topics to provide a comprehensive idea of what online video advertising is to those who are interested in the subject. As the title of this report states, its original purpose is to deliver an introductory guideline to internet video advertising. / text