ETD Collection

Permanent URI for this collectionhttps://wiredspace.wits.ac.za/handle/10539/104


Please note: Digitised content is made available at the best possible quality range, taking into consideration file size and the condition of the original item. These restrictions may sometimes affect the quality of the final published item. For queries regarding content of ETD collection please contact IR specialists by email : IR specialists or Tel : 011 717 4652 / 1954

Follow the link below for important information about Electronic Theses and Dissertations (ETD)

Library Guide about ETD

Browse

Subject: search.filters.subject.Image processing

Search Results

Now showing 1 - 4 of 4
  • No Thumbnail Available
    Item
    SummaryNet: two-stream convolutional networks for automatic video summarisation
    (2020) Jappie, Ziyad
    Video summarisation is the task of automatically summarising a video sequence, to extract “important” parts of the video so as to give an overview of what has occurred. The benefit of solving this problem is that it can be applied to a myriad of fields such as the entertainment industry, sports, e-learning and many more. There is a distinct inherent difficulty with video summarisation due to its subjectivity - there is no one defined correct answer. As such, it is particularly difficult to define and measure tangible performance. This is in addition to the other difficulties associated with general video processing. We present a novel two-stream network framework for automatic video summarisation, which we call SummaryNet. The SummaryNet employs a deep two-stream network to model pertinent spatio-temporal features by leveraging RGB as well as optical flow information. We use the Two-Stream Inflated 3D ConvNet (I3D) network to extract high-level, semantic feature representations as inputs to our SummaryNet model. Experimental results on common benchmark datasets show that the considered method achieves comparable or better results than the state-of-the-art video summarisation methods
  • No Thumbnail Available
    Item
    The forgotten: The ability of the bodyto carry the stories of many
    (2018) Khalishwayo
    Performance has long been a method for telling the stories of people. Through the compassionate engagement of finding the stories of other people, the researcher as performer begins to truthfully realise the stories of others. This study is an exploration of performance as research as a method of finding, holding, unfolding and giving voice to the stories of the marginalised. The creative research, through performance as the written word, performance as the body, performance as ritual theatre, and performance as healing, sought to find ways to give voice to the voiceless. Within this context, this research focussed specifically on the young people’s stories, including this researcher’s story, about being part of FeesMustFall movement, being the subject that gave rise to FeesMustFall movement, and being the forgotten story during the FeesMustFall movement. The creative research sought to grapple with the stories of FeesMustFall movement, to capture the uncaptured stories, to give voice to the voiceless, and to explore through performance as research what the underlying experiences, relationships and psychological condition was that gave rise to FeesMustFall movement, and what the stories have become for those who were deeply impacted by the meaning of the movement and the actions of the movement itself. The research sought to undertake this investigation and present its findings through performance. This part of the creative research serves to reflect on the performance as research and to capture the voices through a performative form of writing. The study seeks to demonstrate the power of performance as a means to knowledge and emancipation. Key Words: FeesMustFall, forgotten stories, performance as research, body, body as performance
  • No Thumbnail Available
    Item
    Mapping illegal dumping using a high resolution remote sensing image case study: Soweto township in South Africa
    (2017) Selani, Lungile
    Although a vast number of illegal dumping investigations have been conducted in the City of Johannesburg by City of Johannesburg Municipality, Government, Corporates as well as NGOs previously, there has been a limited attempt to integrate available datasets from the different methods of illegal dumping monitoring (satellite, spatial data collection and ground-based observations) and GIS modelling. Most South African municipal administrations have had to acknowledge their incapability to cope with the difficulty of illegal dumping monitoring. Illegal dumping challenges often emanate from the incapacity of municipality administrations to meet the required assemblage and removal of wastes. Vacant or unoccupied land is the target of illegal dumping in most areas. This study compares modelled, satellite and collected data using GIS methods to determine the most accurate estimate of detecting illegal dumping. A comparison between Random Forest (RF) and Support Vector Machine (SVM) in mapping illegal dumping and to quantity the significance of Worldview-2 band in detecting and mapping illegal dumping was pursued. Two results were generated: multispectral imagery sorting production using machine-learning RF and SVM algorithms in a comparable land and definition of the significance of unrelated WorldView bands on sorting production. Precision of the derivative thematic maps was evaluated by calculating mix-up milieus of the classifiers’ land use/ land cover maps with separate autonomous justification data sets. A complete classification accurateness of 84.07 % with a kappa value of 0.8116, and 85.16% with a kappa value of 0.8238 was attained using RF and SVM, respectively. An assessment of diverse WorldView-2 bands using the two classifiers indicated that the blend of the red-edge band had a vital consequence on the overall classification accurateness in mapping of illegal dumping. Keywords: Illegal dumping, remote sensing, monitoring, vegetation, spatial datasets, image processing, image classification.
  • No Thumbnail Available
    Item
    The use of anatomically based models for the analysis of imaged tracer experiments in humans
    (1994) Fine, David Robert
    Organ function is often characterised using imaging techniques. In particular a tracer is often used which does not react with tissue, is low in concentration, follows body fluid flows and is distinguishable from the observed system and thus measurable. These requirements ensure linear characteristics of the tracer. In this thesis, these linear characteristics are used to develop a generalised mathematical theory to determine organ function from imaged tracer experiments. The theory is based on anatomical and physiological information for single and multiple input organs. [Abbreviated Abstract. Open document to view full version]