Image recognition for apparel (P3004-131)
In recent years, mobile phone has become smaller and more powerful by its remarkable improvement in hardware. The current smartphone which incorporated the 4G service allows the user to access internet anytime and anywhere, and its incredible built-in camera with high resolution enables the user to...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
2015
|
| Subjects: | |
| Online Access: | http://hdl.handle.net/10356/64204 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | In recent years, mobile phone has become smaller and more powerful by its remarkable improvement in hardware. The current smartphone which incorporated the 4G service allows the user to access internet anytime and anywhere, and its incredible built-in camera with high resolution enables the user to take a clear and beautiful photo, thus, the mobile phone could be used for developing an application for image recognition. And an application for Apparel Recognition was proposed and developed for this project. This project leverages on the image recognition for the apparel (targeted to graphic tees). The purpose of this project is to design a program to identify graphic tee by just taking a picture from magazines, fashion shows or even on the street. And it mainly focuses on creating the database, matching images and outputting the result for further processing. In this project, Bag-of-words model is used in this project, and the steps of feature extraction and representation, codebook construction, histogram representation are discussed in Chapter 2. A reference database that contained 468 reference images for graphic tees (mentioned in Chapter 3.2.3) was created. The database is served as the pool of trained images. Another set of test images is created to test the image recognition program against the reference database. The project makes use of Scale Invariant Feature Transform (SIFT) key point detection method to perform feature matching to find a list of ranked candidate images. The list would be re-ranked by the process of Geometric Verification to find the best match. An additional process is proposed and tested to further improve the recognition rate of the graphic tee, and the accuracy rate that achieved is very promising. |
|---|