Name
Abstract | ||
|---|---|---|
Nowadays, most image classification and recognition problem is nearly solved by deep learning. But some problem is still remaining. To solve remaining problem, many study is still in progress.As part of these study, in this paper, we suggest novel Deep Convolutional Neural Network (DCNN) architecture for face recognition. In order to observe various feature like shape and texture from face, we design different size filters (3x3, 5x5, and 7x7). After convolution step, we merge these filters and distribute to various size convolution layer again. This deep convolutional network is constructed by 3x3 convolution layer, 3x3 max pooling layer, and 2 fully-connected layer. Our architecture is bigger than ordinary network architecture, but it is more efficient than ensemble of independent three networks.Our method uses parallel convolution layer what is calculated by various size filters. It looks like large neural network and it brings about positive effect like ensemble models. |
Year | Venue | Keywords |
|---|---|---|
2016 | 2016 13TH INTERNATIONAL CONFERENCE ON UBIQUITOUS ROBOTS AND AMBIENT INTELLIGENCE (URAI) | Convolutional Neural Network, Face Recognition, Deep Learning |
Field | DocType | ISSN |
Facial recognition system,Pattern recognition,Computer science,Convolutional neural network,Artificial intelligence,Deep learning,Contextual image classification | Conference | 2325-033X |
Citations | PageRank | References |
0 | 0.34 | 7 |
Authors | ||
3 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Jaeyoon Jang | 1 | 0 | 0.34 |
| Young-Jo Cho | 2 | 149 | 20.92 |
| Ho-Sub Yoon | 3 | 11 | 3.88 |