| Author: | Pan, Yikun |
| Title: | Image segmentation and inpainting based on deep learning |
| Advisors: | Lun, Daniel (EIE) |
| Degree: | M.Sc. |
| Year: | 2020 |
| Subject: | Image analysis Image processing Machine learning Hong Kong Polytechnic University -- Dissertations |
| Department: | Department of Electronic and Information Engineering |
| Pages: | 53 pages : color illustrations |
| Language: | English |
| Abstract: | Based on the deep learning methods, many intelligent image processing algorithms are employed in the field of computer vision in recent years. One of them is image segmentation, which is a commonly used technique in digital image processing for partitioning an image into multiple parts or regions based on the characteristics of the pixels in the image. Via learning the image features from batches of training samples, the convolutional neural network has become one of the powerful tools in the spatially dense prediction tasks of semantic segmentation. To make the network keep the global contextual information of the whole image, a pyramid pooling module is used in this project to accomplish different region-based context aggregation. Another application that the deep learning methods have been successfully applied is image inpainting, which aims at reconstructing the missing or corrupted regions within a picture or an image. For image inpainting, the Generative Adversarial Network (GAN) has been commonly used for generating the predicted outputs with a synthesis model trained based on the competition between a generator and a discriminator. For example, the two-stage adversarial model of the method EdgeConnect comprises an edge generator followed by an image completion network. It is adopted in this project for reproducing the missing regions exhibiting fine details. In this project, an entertaining application is developed by using image segmentation and an inpainting model that can interact with the users for removing the unwanted object in an image and filling with meaningful content. Users can mask out the unwanted object in an image through the segmentation model and fill the masked region with meaningful content through the inpainting model. During the stage of inpainting, users are allowed to sketch the outline of the objects inside the masked region. After receiving the input from the user, the system performs much better and can give the desired results that the user wants. |
| Rights: | All rights reserved |
| Access: | restricted access |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| 5169.pdf | For All Users (off-campus access for PolyU Staff & Students only) | 3.35 MB | Adobe PDF | View/Open |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/10767