Deep learning
Kawalya, Davis (2022)
Kawalya, Davis
2022
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-202205118604
https://urn.fi/URN:NBN:fi:amk-202205118604
Tiivistelmä
This thesis builds upon work carried out by the author of this thesis recently on deep learning to build a computer vision deep learning pipeline using Generative Adversarial Networks (GANs). For much of the work, agile development methodologies were employed for each task in a workflow that moved from to-do then to research, study, analysis then to design, then to implementation then to testing then to done/completed done. The computer vision deep learning pipeline has many use-cases to which it can be applied as will be seen later when the tools used to set it up and the results on some use-cases achieving 86% inference accuracy on aircraft recognition and 90% inference accuracy (95% AUC) on disease recognition (lateral projection data) and 86% inference accuracy (93% AUC) on disease recognition (four projection data) are discussed. While humans may be able to achieve such inference accuracies on relatively small datasets, deep learning is able to achieve the inference accuracies much faster be it on small or on very large datasets. Other systems (documentation and Constructive Cost Models (COCOMO) costs provided as attachments) that have been developed during this thesis have been deemed to be beyond the scope of this thesis due to the complexity they present in their design and implementation.