- BDU IR Home
- →
- Bahir Dar Institute of Technology (BiT)
- →
- Faculty of Computing
- →
- Computer Science
- →
- thesis
- →
- View Item
JavaScript is disabled for your browser. Some features of this site may not work without it.
| dc.contributor.author | SELAMAWIT, WORKINEH | |
| dc.date.accessioned | 2022-03-18T06:43:05Z | |
| dc.date.available | 2022-03-18T06:43:05Z | |
| dc.date.issued | 2021-09-13 | |
| dc.identifier.uri | http://ir.bdu.edu.et/handle/123456789/13213 | |
| dc.description.abstract | The Ethiopia Orthodox Tewahedo Church (EOTC) has its own icons the people used for reverence. Currently, the Ethiopian Orthodox Tewahedo Church (EOTC) icons have mixed with different western country icons due to the advancement of globalization and influences of western cultures. In the previous several types of research can be done in the area of icon classification and recognition by using image processing techniques. However, the Ethiopian Orthodox Tewahedo Church (EOTC) icon detection is still a problem for the people who do not know about the Ethiopia Orthodox Church dogma and doctrine. Most of the previous work focuses on humans like face appearance icons based on Byzantine’s icon canon by considering only texture, shape, and color. They used the handcrafted feature extraction techniques only. In this thesis, we developed a model using a combination of the Gabor feature, Local feature, and Conventional Neural Networks (CNN) together to extract the deep feature and the local feature of the Ethiopian Orthodox Tewahedo Church icon by considering the most relevant feature of EOTC icon like shape, position, angle of orientation, texture, and direction. To detect whether the icon is Ethiopic or Non-Ethiopic we used Support Vector Machine (SVM). The datasets are collected in the ancient Ethiopian Orthodox Tewahedo Church (EOTC and markets that are found around Bahir Dar Ethiopia. We collected 746 Ethiopic icon images and 721 Non-Ethiopic icon images with 512x512 image size. From the 1467 icon images 80% dataset to train the model and 20% dataset to evaluate the performance of the model. Finally to evaluate the performance of the model we used a Support Vector Machine (SVM) classifier. After evaluating the model we achieved 90.06%, 86.73%, 84.69%, 85.03%, 92.51%, 93.54%, and 95.24% accuracy for end-to-end CNN and CNN, HOG, and Gabor, CNN_HOG, CNN_Gabor, and CNN_HOG_Gabor with an SVM classifier respectively. Keywords: EOTC, Icon, Computer Vision, Gabor feature and local feature, CNN | en_US |
| dc.language.iso | en_US | en_US |
| dc.subject | computer science | en_US |
| dc.title | GABOR FEATURE AND LOCAL FEATURE ASSISTED CNN FOR ETHIOPIAN ORTHODOX TEWAHEDO CHURCH ICON DETECTION | en_US |
| dc.type | Thesis | en_US |
