Abstract
We propose a local modelling approach using deep convolutional neural networks (CNNs) for fine-grained image classification. Recently, deep CNNs trained from large datasets have considerably improved the performance of object recognition. However, to date there has been limited work using these deep CNNs as local feature extractors. This partly stems from CNNs having internal representations which are high dimensional, thereby making such representations difficult to model using stochastic models. To overcome this issue, we propose to reduce the dimensionality of one of the internal fully connected layers, in conjunction with layer-restricted retraining to avoid retraining the entire network. The distribution of low-dimensional features obtained from the modified layer is then modelled using a Gaussian mixture model. Comparative experiments show that considerable performance improvements can be achieved on the challenging Fish and UEC FOOD-100 datasets.
Original language | English |
---|---|
Title of host publication | 2015 IEEE International Conference on Image Processing (ICIP 2015) |
Editors | Fabrice Labeau, Jean-Philippe Thiran |
Place of Publication | Piscataway NJ USA |
Publisher | IEEE, Institute of Electrical and Electronics Engineers |
Pages | 4112-4116 |
Number of pages | 5 |
ISBN (Electronic) | 9781479983391 |
ISBN (Print) | 9781479983407 |
DOIs | |
Publication status | Published - 9 Dec 2015 |
Externally published | Yes |
Event | IEEE International Conference on Image Processing 2015 - Quebec City, Canada Duration: 27 Sept 2015 → 30 Sept 2015 Conference number: 22nd http://icip2015.org/index.html https://ieeexplore.ieee.org/xpl/conhome/7328364/proceeding (Proceedings) |
Conference
Conference | IEEE International Conference on Image Processing 2015 |
---|---|
Abbreviated title | ICIP 2015 |
Country/Territory | Canada |
City | Quebec City |
Period | 27/09/15 → 30/09/15 |
Internet address |
Keywords
- deep convolutional neural networks
- fine-grained classification
- Gaussian mixture models
- session variation modelling