Encapsulating the Impact of Transfer Learning, Domain Knowledge and Training Strategies in Deep-Learning Based Architecture: A Biometric Based Case Study

Avantika Singh, Aditya Nigam; The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2018, pp. 1866-1868

Abstract


In this paper, efforts have been made to analyze the impact of training strategies, transfer learning and domain knowledge on two biometric-based problems namely: three class oculus classification and fingerprint sensor classification. For analyzing these problems we have considered deep-learning based architecture and evaluated our results on benchmark contact-lens datasets like IIIT-D, ND, IIT-K (our model is publicly available) and on fingerprint datasets like FVC-2002, FVC-2004, FVC-2006, IIITD-MOLF. In-depth feature analysis of various proposed deep-learning models has been done in order to infer that indeed training in different ways along with transfer learning and domain knowledge plays a vital role in deciding the learning ability of any network.

Related Material


[pdf]
[bibtex]
@InProceedings{Singh_2018_CVPR_Workshops,
author = {Singh, Avantika and Nigam, Aditya},
title = {Encapsulating the Impact of Transfer Learning, Domain Knowledge and Training Strategies in Deep-Learning Based Architecture: A Biometric Based Case Study},
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
month = {June},
year = {2018}
}