Javascript must be enabled for the correct page display

Data Set Extension with Generative Adversarial Nets

Boulogne, Luuk (2018) Data Set Extension with Generative Adversarial Nets. Master's Thesis / Essay, Artificial Intelligence.


Download (3MB) | Preview
[img] Text
Restricted to Registered users only

Download (98kB)


This thesis focuses on supplementing data sets with data of absent classes by using other, similar data sets in which these classes are represented. The data is generated using Generative Adversarial Nets (GANs) trained on the CelebA and MNIST data sets. In particular, this thesis involves Coupled GANs (CoGANs), Auxiliary Classifier GANs (AC-GANs) and a novel combination of the two, Coupled Auxiliary Classifier GANs (CoAC-GANs). The abilities of these GANs to generate image data of domain-class combinations that were removed from the training data are compared. Classifiers are trained on the generated data to investigate the usefulness of the generated data for data set extension. The results show that AC-GANs and CoAC-GANs can be used successfully to generate labeled data from domain-class combinations that are absent from the training data. Furthermore, they suggest that the preference for one of the two types of generative models depends on training set characteristics. Classifiers trained on the generated data can accurately classify unseen data from the missing domain-class combinations.

Item Type: Thesis (Master's Thesis / Essay)
Supervisor name: Wiering, M.A. and Dijkstra, K.
Degree programme: Artificial Intelligence
Thesis type: Master's Thesis / Essay
Language: English
Date Deposited: 20 Jul 2018
Last Modified: 23 Jul 2018 13:05

Actions (login required)

View Item View Item