Sorry, you need to enable JavaScript to visit this website.

facebooktwittermailshare

CONTRASTIVE-CENTER LOSS FOR DEEP NEURAL NETWORKS

Abstract: 

The deep convolutional neural network(CNN) has significantly raised the performance of image classification and face
recognition. Softmax is usually used as supervision, but it only penalizes the classification loss. In this paper, we propose a novel auxiliary supervision signal called contrastive-center loss, which can further enhance the discriminative power of the features, for it learns a class center for each class. The proposed contrastive-center loss simultaneously considers intra-class compactness and inter-class separability, by penalizing the contrastive values between: (1)the distances of training samples to their corresponding class centers, and (2)the sum of the distances of training samples to their non-corresponding class centers. Experiments on different datasets demonstrate the effectiveness of contrastive-center loss.

up
0 users have voted:

Paper Details

Authors:
Submitted On:
13 September 2017 - 10:46pm
Short Link:
Type:
Poster
Event:
Presenter's Name:
Ce Qi
Paper Code:
1592
Document Year:
2017
Cite

Document Files

poster

(0)

Subscribe

[1] , "CONTRASTIVE-CENTER LOSS FOR DEEP NEURAL NETWORKS", IEEE SigPort, 2017. [Online]. Available: http://sigport.org/1990. Accessed: Nov. 20, 2017.
@article{1990-17,
url = {http://sigport.org/1990},
author = { },
publisher = {IEEE SigPort},
title = {CONTRASTIVE-CENTER LOSS FOR DEEP NEURAL NETWORKS},
year = {2017} }
TY - EJOUR
T1 - CONTRASTIVE-CENTER LOSS FOR DEEP NEURAL NETWORKS
AU -
PY - 2017
PB - IEEE SigPort
UR - http://sigport.org/1990
ER -
. (2017). CONTRASTIVE-CENTER LOSS FOR DEEP NEURAL NETWORKS. IEEE SigPort. http://sigport.org/1990
, 2017. CONTRASTIVE-CENTER LOSS FOR DEEP NEURAL NETWORKS. Available at: http://sigport.org/1990.
. (2017). "CONTRASTIVE-CENTER LOSS FOR DEEP NEURAL NETWORKS." Web.
1. . CONTRASTIVE-CENTER LOSS FOR DEEP NEURAL NETWORKS [Internet]. IEEE SigPort; 2017. Available from : http://sigport.org/1990