Uncertainty-Handling of Multi-Object Categorization in Scenes using Fuzzy Swarm Intelligence-based Deep-Belief Network
S. Kumaravel1, S. Veni2
1S. Kumaravel, Research Scholar, PhD, PT Department of Computer Science, Karpagam Academy of Higher Education, Coimbatore (Tamil Nadu), India.
2Dr. S. Veni, Professor & HOD, Department of Computer Science, Karpagam Academy of Higher Education, Coimbatore (Tamil Nadu), India.
Manuscript received on 25 May 2019 | Revised Manuscript received on 03 June 2019 | Manuscript Published on 22 June 2019 | PP: 63-72 | Volume-8 Issue-3S, February 2019 | Retrieval Number: C10140283S19/19©BEIESP
Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: In current digital era, most people are using multimedia not only for entertainment, but also for various commercial usages. As the demand for multimedia increased vigorously, the researchers have started focusing on machine learning systems, in image processing. Traditional models outfit the process of image recognition, classification, etc., due to their excessive size and computation time. To overcome the aforementioned issues in conventional classification models, deep learning greatly influences the researchers in the field of image processing, with vast amount of qualities. This paper aims at developing an optimized fuzzy swarm intelligence-based deepbelief network (FSIDBN), which handles the issue of multiple object categorization in scenes. The uncertainty in handling is one of the prominent issues in image processing, when there is vagueness in determining, a greater number of objects. At the same time, conventional deep-belief network itself holds some of the disadvantages, when it assigns weights randomly. The existing issues are overcome, by representing the input images in fuzzy domain using degree of membership, to which each object belongs, and assigning optimal weights on each layer of stacked Restricted Boltzmann Machines using Fuzzy Swarm intelligence. The simulation result proves that, FSIDBN has achieved a higher degree of accuracy in the categorization of multiple distinct objects, compared to FDBN and DBN models.
Keywords: Multi-Object Categorization, Fuzzy Deep Belief Network, Classification, Membership and Restricted Boltzmann Machine.
Scope of the Article: Embedded Networks