Human Image Recognition Design through Neural Network Model
| Autor: | LAI, LI-YUNG, 賴麗雲 |
|---|---|
| Rok vydání: | 2018 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 106 This thesis develops the higher accuracy Neural Networks image model system based on the image processing and deep learning concepts. The restricted boltzmann machine (RBM) is considered as a basic unit to construct the multilayers Neural Networks system structure. Firstly, an unsupervised learning pre-training and Deep Belief Nets (DBN) approach the initial Convolutional-Neural Networks (CNN) system structure. A backpropagation learning algorithm is proposed to complete all weights tuning in the deep-learning neural networks. After learning iteration, the deep neural networks is contained the extracted feature of trading dataset and construct the required parameters of the human body recognition system. This study has finished (1) human feature image processing (2) deep-learning neural-networks in functional approximation and (3) human body pose and motion estimation experiments. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|