Advanced

Transferring and compressing convolutional neural networks for face representations

Grundström, Jakob; Chen, Jiandan LU ; Ljungqvist, Martin Georg and Åström, Kalle LU (2016) 13th International Conference on Image Analysis and Recognition, ICIAR 2016 In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 9730. p.20-29
Abstract

In this work we have investigated face verification based on deep representations from Convolutional Neural Networks (CNNs) to find an accurate and compact face descriptor trained only on a restricted amount of face image data. Transfer learning by fine-tuning CNNs pre-trained on large-scale object recognition has been shown to be a suitable approach to counter a limited amount of target domain data. Using model compression we reduced the model complexity without significant loss in accuracy and made the feature extraction more feasible for real-time use and deployment on embedded systems and mobile devices. The compression resulted in a 9-fold reduction in number of parameters and a 5-fold speed-up in the average feature extraction... (More)

In this work we have investigated face verification based on deep representations from Convolutional Neural Networks (CNNs) to find an accurate and compact face descriptor trained only on a restricted amount of face image data. Transfer learning by fine-tuning CNNs pre-trained on large-scale object recognition has been shown to be a suitable approach to counter a limited amount of target domain data. Using model compression we reduced the model complexity without significant loss in accuracy and made the feature extraction more feasible for real-time use and deployment on embedded systems and mobile devices. The compression resulted in a 9-fold reduction in number of parameters and a 5-fold speed-up in the average feature extraction time running on a desktop CPU. With continued training of the compressed model using a Siamese Network setup, it outperformed the larger model.

(Less)
Please use this url to cite or link to this publication:
author
organization
publishing date
type
Chapter in Book/Report/Conference proceeding
publication status
published
subject
in
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
volume
9730
pages
10 pages
publisher
Springer Verlag
conference name
13th International Conference on Image Analysis and Recognition, ICIAR 2016
external identifiers
  • Scopus:84978872481
ISSN
0302-9743
1611-3349
ISBN
978-3-319-41500-0
DOI
10.1007/978-3-319-41501-7_3
language
English
LU publication?
yes
id
04a61dac-b616-4585-8046-f35e569f7933
date added to LUP
2016-08-29 08:37:06
date last changed
2017-02-16 11:15:53
@inproceedings{04a61dac-b616-4585-8046-f35e569f7933,
  abstract     = {<p>In this work we have investigated face verification based on deep representations from Convolutional Neural Networks (CNNs) to find an accurate and compact face descriptor trained only on a restricted amount of face image data. Transfer learning by fine-tuning CNNs pre-trained on large-scale object recognition has been shown to be a suitable approach to counter a limited amount of target domain data. Using model compression we reduced the model complexity without significant loss in accuracy and made the feature extraction more feasible for real-time use and deployment on embedded systems and mobile devices. The compression resulted in a 9-fold reduction in number of parameters and a 5-fold speed-up in the average feature extraction time running on a desktop CPU. With continued training of the compressed model using a Siamese Network setup, it outperformed the larger model.</p>},
  author       = {Grundström, Jakob and Chen, Jiandan and Ljungqvist, Martin Georg and Åström, Kalle},
  booktitle    = {Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)},
  isbn         = {978-3-319-41500-0},
  issn         = {0302-9743},
  language     = {eng},
  pages        = {20--29},
  publisher    = {Springer Verlag},
  title        = {Transferring and compressing convolutional neural networks for face representations},
  url          = {http://dx.doi.org/10.1007/978-3-319-41501-7_3},
  volume       = {9730},
  year         = {2016},
}