Loading...

Realization Law of Pragnanz and Closure of Gestalt Theory Using Neural Network Modeling and Active Learning Method

Safaei, Negin | 2017

864 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: English
  3. Document No: 49674 (55)
  4. University: Sharif University of Technology, International Campus, Kish Island
  5. Department: Science and Engineering
  6. Advisor(s): Bagheri Shouraki, Saeed; Mirian Hossinabadi, Maryam
  7. Abstract:
  8. Human brain is one the most efficient biological neural network that is capable of performing most complicated tasks in very short time. Visual system as a part of the brain can process the visual records from the environment and perform very useful processes on them. The information provided from visual system processes play important roles in human intelligence. In this thesis we introduce and investigate one of the well-known theories about visual system called gestalt theories. These theories tried to define how a biological neural network can perform such results. Gestalt theorists developed rules of perception to explain their ideas including Law of Pragnanz, Similarity, Closure, proximity etc. This thesis presents a new algorithm for predicting missing parts of images. Active learning method (ALM) is used as an artificial intelligent approach to realize law of Pragnanz and Closure principle of Gestalt theory. ALM is a pattern-based algorithm for soft computing which uses the Ink Drop Spread (IDS) algorithm. Our framework learns the patterns in an unsupervised manner or alongside any supervised task according to symmetry and pattern of rest of shape. This method inspired by Morphology and used ALM as an iterative process. At each iteration proposed algorithm find Center of Gravity (COG) and we can achieve the accuracy requirement based on distances between centers to predict removed segment of image. Also this law realized by the other attitude which supervised learning and used Convolutional Neural Networks
  9. Keywords:
  10. Morphology ; Active Learning ; Convolutional Neural Network ; Ink Drop Spread (IDS)Operator ; Basis Function ; Pragnanz Law ; Closure Law ; Gestalt Theory

 Digital Object List

 Bookmark

No TOC