Fine-grained artworks classification

In this thesis, we apply deep convolutional neural networks to ne-grained artwork classification on the large-scale painting collection, WikiArt. We propose a new architecture that aggregates features from different convolutional layers to exploit earlier layer features. The new architecture is evaluated on the challenging fine-grained artist and year classification. We also propose a regularization method that penalizes correlations of convolutional feature maps. With the decorrelation regularization, we further improve the classification accuracy of the proposed architecture.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Computer Science

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2018

Author dc:creator

• Huang, Jing

Contributors dc:contributor

• Lazebnik, Svetlana

• Convolutional Neural Networks
• Fine-grained Classification
• Artworks Classification

Statement dc:rights

• Copyright 2018 Jing Huang

Language dc:language

en

Handle dc:identifier

http://hdl.handle.net/2142/101050

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/101050