Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor | Faculty of Engineering | en_US |
| dc.contributor.advisor | Chi, Zheru George (EIE) | - |
| dc.creator | Lu, Chongkai | - |
| dc.identifier.uri | https://theses.lib.polyu.edu.hk/handle/200/10097 | - |
| dc.language | English | en_US |
| dc.publisher | Hong Kong Polytechnic University | - |
| dc.rights | All rights reserved | en_US |
| dc.title | A study on spatial transformer networks | en_US |
| dcterms.abstract | Neural networks have already shown their excellent performance on image recognition, while most of them perform well only when test samples are aligned well. However, in real world, most of the targets required detection and recognition in images are various in size, rotation and location. Spatial Transformer Networks (STN) as an end-to-end neural network which combine learning alignment and classification into one neural network, can learn to align targets automatically in cooperation with a classification neural network. STN shows good performance on image recognition on spatially variant datasets. We implemented STN and found the averaging problem of STN. We propose a new training strategy for STN called STP-NN (Spatial Transformer Pretrained-Neural Networks). Experimental results on MNIST shows that STP-NN can achieve a better performance than normal STN, decreasing the error rate on spatially the distorted MNIST dataset from 2.75% to 1.86%. Moreover, shown in our experimental results, STP-NN can align the rotated images well but STN shows a bad performance on these images. | en_US |
| dcterms.extent | xi, 64 pages : color illustrations | en_US |
| dcterms.isPartOf | PolyU Electronic Theses | en_US |
| dcterms.issued | 2019 | en_US |
| dcterms.educationalLevel | M.Sc. | en_US |
| dcterms.educationalLevel | All Master | en_US |
| dcterms.LCSH | Hong Kong Polytechnic University -- Dissertations | en_US |
| dcterms.LCSH | Image processing | en_US |
| dcterms.LCSH | Neural networks (Computer science) | en_US |
| dcterms.LCSH | Machine learning | en_US |
| dcterms.accessRights | restricted access | en_US |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| 991022270856703411.pdf | For All Users (off-campus access for PolyU Staff & Students only) | 1.35 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/200/10097