Document Details
Document Type |
: |
Thesis |
Document Title |
: |
AUTOMATIC IMAGE ANNOTATION BASED ON GENERATIVE ADVERSARIAL NETWORKS التعليق التلقائي للصور باستخدام الشبكة الخصومية التوليدية |
Subject |
: |
Faculty of Computing and Information Technology |
Document Language |
: |
Arabic |
Abstract |
: |
Nowadays, Automatic Image Annotation (AIA) has been adopted in different applications such as image retrieval and classification. Indeed, AIA plays a significant role in enhancing image retrieval. Deep Learning is used in AIA to extract image features and then convert these features into text descriptions (i.e., labels). It is currently considered one of the most attractive research areas. However, Conventional AIA models that employ deep learning methods suffer from various shortcomings, such as poor annotation performance. Accordingly, this thesis proposes an AIA model based on convolutional neural networks (CNNs), Generative Adversarial Networks (GANs), and transfer learning. The proposed model aims to overcome the limitations of the existing models such as the challenge of training a CNN model using small-sized datasets that affect the effectiveness of AIA. GANs have attracted a lot of interest in the computer vision field because of its ability to generate data without explicitly using probability density. Thus, it has proven its usefulness in image annotation, and image augmentation. In this research, an Auxiliary classifier-GAN (ACGAN) has been used, where the discriminator predicts the class of an image rather than taking it as a given input; therefore, the stabilization of the training stage is ensured, and the generation of high-quality images is provided. Also, transfer learning is used to enhance the performance of the classification. The proposed model enhanced the best-scored state-of-art model, using ImageClef dataset, by 5.61% in terms of MiAP, a gain of 16.15% in terms of F-measure, and achieved a 69.8% reduction in terms of EER. Also, using ESPGame dataset, the proposed model enhanced the best-scored state-of-art model by 130% in terms of F-measure. Last, using the IAPR-TC12 dataset, the proposed model improved to the best-scored state-of-art model, by 119% in terms of F-measure. |
Supervisor |
: |
Dr. Mounira Taileb |
Thesis Type |
: |
Master Thesis |
Publishing Year |
: |
1443 AH
2022 AD |
Co-Supervisor |
: |
Dr. Reem Alotaibi |
Added Date |
: |
Tuesday, March 15, 2022 |
|
Researchers
عبير محمد الشهري | Alshehri, Abeer Mohammed | Researcher | Master | |
|
Back To Researches Page
|