Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Face Recognition Based on Embedded Systems

Björgvinsdottir, Hanna Bara LU and Seibold, Robin LU (2016) In Master's Theses in Mathematical Sciences FMA820 20161
Mathematics (Faculty of Engineering)
Abstract
Machine learning in general, and artificial neural networks in particular, have gained a lot of attention in recent years. Using deep neural networks for classification tasks, such as face recognition, has proven more and more successful over time. The performance increase is partly due to more complex network architectures, and partly due to the use of larger datasets.
The increased complexity of networks has lead to an increase in parameters, which in turn results in slower training and inference, making it hard to deploy such models on limited hardware.
The main objective of this master's thesis is to train a convolutional neural network for face recognition, and deploy it on an embedded system, with the aim of real-time performance.
... (More)
Machine learning in general, and artificial neural networks in particular, have gained a lot of attention in recent years. Using deep neural networks for classification tasks, such as face recognition, has proven more and more successful over time. The performance increase is partly due to more complex network architectures, and partly due to the use of larger datasets.
The increased complexity of networks has lead to an increase in parameters, which in turn results in slower training and inference, making it hard to deploy such models on limited hardware.
The main objective of this master's thesis is to train a convolutional neural network for face recognition, and deploy it on an embedded system, with the aim of real-time performance.
By using transfer-learning as a means to adjust a pre-trained model to fit new data, the time needed for the training phase is reduced. The resulting model achieves an accuracy of 91.66%, while distinguishing between 2,904 identities.
The model is then compressed by a method referred to as pruning, reducing the amount of parameters in the fully connected layers by a factor of 20, greatly reducing the memory footprint while remaining within 1% of the original accuracy.
Finally, by combining the resulting neural network model with a custom built framework and a live video stream, real-time face recognition is achieved on an embedded device. (Less)
Popular Abstract (Swedish)
Neurala nätverk har visat större och större framsteg de senaste åren. Framstegen beror till stor del på nätverkens alltmer komplexa arkitektur, vilket medför att kraven som ställs på hårdvaran som nätverken ska användas på också ökar. Målet med detta examensarbete är att träna ett neuralt nätverk för ansiktsigenkänning, och använda det på ett inbyggt system med begränsad hårdvara.
Please use this url to cite or link to this publication:
author
Björgvinsdottir, Hanna Bara LU and Seibold, Robin LU
supervisor
organization
alternative title
Ansiktsigenkänning För Inbyggda System
course
FMA820 20161
year
type
H2 - Master's Degree (Two Years)
subject
keywords
Deep learning, neural networks, face recognition, embedded systems
publication/series
Master's Theses in Mathematical Sciences
report number
LUTFMA-3293-2016
ISSN
1404-6342
other publication id
2016:E17
language
English
id
8883824
date added to LUP
2016-08-25 13:48:49
date last changed
2016-08-25 13:48:49
@misc{8883824,
  abstract     = {{Machine learning in general, and artificial neural networks in particular, have gained a lot of attention in recent years. Using deep neural networks for classification tasks, such as face recognition, has proven more and more successful over time. The performance increase is partly due to more complex network architectures, and partly due to the use of larger datasets.
The increased complexity of networks has lead to an increase in parameters, which in turn results in slower training and inference, making it hard to deploy such models on limited hardware.
The main objective of this master's thesis is to train a convolutional neural network for face recognition, and deploy it on an embedded system, with the aim of real-time performance.
By using transfer-learning as a means to adjust a pre-trained model to fit new data, the time needed for the training phase is reduced. The resulting model achieves an accuracy of 91.66%, while distinguishing between 2,904 identities.
The model is then compressed by a method referred to as pruning, reducing the amount of parameters in the fully connected layers by a factor of 20, greatly reducing the memory footprint while remaining within 1% of the original accuracy.
Finally, by combining the resulting neural network model with a custom built framework and a live video stream, real-time face recognition is achieved on an embedded device.}},
  author       = {{Björgvinsdottir, Hanna Bara and Seibold, Robin}},
  issn         = {{1404-6342}},
  language     = {{eng}},
  note         = {{Student Paper}},
  series       = {{Master's Theses in Mathematical Sciences}},
  title        = {{Face Recognition Based on Embedded Systems}},
  year         = {{2016}},
}