Abstract
We present a novel deep convolutional neural network (DCNN) system for fine-grained image classification, called a mixture of DCNNs (MixDCNN). The fine-grained image classification problem is characterised by large intra-class variations and small inter-class variations. To overcome these problems our proposed MixDCNN system partitions images into K subsets of similar images and learns an expert DCNN for each subset. The output from each of the K DCNNs is combined to form a single classification decision. In contrast to previous techniques, we provide a formulation to perform joint end-to-end training of the K DCNNs simultaneously. Extensive experiments, on three datasets using two network structures (AlexNet and GoogLeNet), show that the proposed MixDCNN system consistently outperforms other methods. It provides a relative improvement of 12.7% and achieves state-of-the-art results on two datasets.
| Original language | English |
|---|---|
| Title of host publication | 2016 IEEE Winter Conference on Applications of Computer Vision (WACV 2016) |
| Editors | Greg Mori, Robert Pless, Scott McCloskey, Rahul Sukthankar |
| Place of Publication | Piscataway NJ USA |
| Publisher | IEEE, Institute of Electrical and Electronics Engineers |
| ISBN (Electronic) | 9781509006410 |
| ISBN (Print) | 9781509006427 |
| DOIs | |
| Publication status | Published - 23 May 2016 |
| Externally published | Yes |
| Event | IEEE Winter Conference on Applications of Computer Vision 2016 - Lake Placid, United States of America Duration: 7 Mar 2016 → 9 Mar 2016 Conference number: 5th http://www.wacv16.org/ http://www.wikicfp.com/cfp/servlet/event.showcfp?eventid=46372©ownerid=26710 http://wacv16.wacv.net/ |
Conference
| Conference | IEEE Winter Conference on Applications of Computer Vision 2016 |
|---|---|
| Abbreviated title | WACV 2016 |
| Country | United States of America |
| City | Lake Placid |
| Period | 7/03/16 → 9/03/16 |
| Internet address |
Cite this
}
Fine-grained classification via mixture of deep convolutional neural networks. / Ge, Zongyuan; Bewley, Alex; McCool, Christopher; Corke, Peter; Upcroft, Ben; Sanderson, Conrad.
2016 IEEE Winter Conference on Applications of Computer Vision (WACV 2016). ed. / Greg Mori; Robert Pless; Scott McCloskey; Rahul Sukthankar. Piscataway NJ USA : IEEE, Institute of Electrical and Electronics Engineers, 2016. 7477700.Research output: Chapter in Book/Report/Conference proceeding › Conference Paper › Research › peer-review
TY - GEN
T1 - Fine-grained classification via mixture of deep convolutional neural networks
AU - Ge, Zongyuan
AU - Bewley, Alex
AU - McCool, Christopher
AU - Corke, Peter
AU - Upcroft, Ben
AU - Sanderson, Conrad
PY - 2016/5/23
Y1 - 2016/5/23
N2 - We present a novel deep convolutional neural network (DCNN) system for fine-grained image classification, called a mixture of DCNNs (MixDCNN). The fine-grained image classification problem is characterised by large intra-class variations and small inter-class variations. To overcome these problems our proposed MixDCNN system partitions images into K subsets of similar images and learns an expert DCNN for each subset. The output from each of the K DCNNs is combined to form a single classification decision. In contrast to previous techniques, we provide a formulation to perform joint end-to-end training of the K DCNNs simultaneously. Extensive experiments, on three datasets using two network structures (AlexNet and GoogLeNet), show that the proposed MixDCNN system consistently outperforms other methods. It provides a relative improvement of 12.7% and achieves state-of-the-art results on two datasets.
AB - We present a novel deep convolutional neural network (DCNN) system for fine-grained image classification, called a mixture of DCNNs (MixDCNN). The fine-grained image classification problem is characterised by large intra-class variations and small inter-class variations. To overcome these problems our proposed MixDCNN system partitions images into K subsets of similar images and learns an expert DCNN for each subset. The output from each of the K DCNNs is combined to form a single classification decision. In contrast to previous techniques, we provide a formulation to perform joint end-to-end training of the K DCNNs simultaneously. Extensive experiments, on three datasets using two network structures (AlexNet and GoogLeNet), show that the proposed MixDCNN system consistently outperforms other methods. It provides a relative improvement of 12.7% and achieves state-of-the-art results on two datasets.
UR - http://www.scopus.com/inward/record.url?scp=84977646599&partnerID=8YFLogxK
U2 - 10.1109/WACV.2016.7477700
DO - 10.1109/WACV.2016.7477700
M3 - Conference Paper
SN - 9781509006427
BT - 2016 IEEE Winter Conference on Applications of Computer Vision (WACV 2016)
A2 - Mori, Greg
A2 - Pless, Robert
A2 - McCloskey, Scott
A2 - Sukthankar, Rahul
PB - IEEE, Institute of Electrical and Electronics Engineers
CY - Piscataway NJ USA
ER -