<div class="csl-bib-body">
<div class="csl-entry">Haas, B., Wendt, A., Jantsch, A., & Wess, M. (2021). Neural Network Compression Through Shunt Connections and Knowledge Distillation for Semantic Segmentation Problems. In I. Maglogiannis, J. MacIntyre, & L. Iliadis (Eds.), <i>Artificial Intelligence Applications and Innovations - 17th IFIP WG 12.5 International Conference, AIAI 2021, Hersonissos, Crete, Greece, June 25–27, 2021, Proceedings</i> (pp. 349–361). Springer Nature Switzerland AG. https://doi.org/10.1007/978-3-030-79150-6_28</div>
</div>
-
dc.identifier.isbn
9783030791490
-
dc.identifier.isbn
9783030791506
-
dc.identifier.uri
http://hdl.handle.net/20.500.12708/77394
-
dc.description.abstract
Employing convolutional neural network models for large
scale datasets represents a big challenge. Especially embedded devices
with limited resources cannot run most state-of-the-art model architectures
in real-time, necessary for many applications. This paper proves
the applicability of shunt connections on large scale datasets and narrows
this computational gap. Shunt connections is a proposed method
for MobileNet compression. We are the first to provide results of shunt
connections for the MobileNetV3 model and for segmentation tasks on
the Cityscapes dataset, using the DeeplabV3 architecture, on which we
achieve compression by 28%, while observing a 3.52 drop in mIoU. The
training of shunt-inserted models are optimized through knowledge distillation.
The full code used for this work will be available online.
en
dc.language.iso
en
-
dc.publisher
Springer Nature Switzerland AG
-
dc.relation.ispartofseries
IFIP Advances in Information and Communication Technology
-
dc.subject
Optimization
-
dc.subject
Accuracy
-
dc.subject
Machine learning
-
dc.subject
Shunt connections
-
dc.subject
Knowledge distillation
-
dc.subject
Latency
-
dc.subject
CIFAR
-
dc.subject
Cityscapes
-
dc.subject
DeepLab
-
dc.subject
MobileNet
-
dc.subject
Embedded machine learning
-
dc.title
Neural Network Compression Through Shunt Connections and Knowledge Distillation for Semantic Segmentation Problems
en
dc.type
Konferenzbeitrag
de
dc.type
Inproceedings
en
dc.contributor.editoraffiliation
University of Piraeus, Greece
-
dc.contributor.editoraffiliation
University of Sunderland, United Kingdom of Great Britain and Northern Ireland (the)
-
dc.relation.isbn
978-3-030-79149-0
-
dc.relation.doi
10.1007/978-3-030-79150-6
-
dc.relation.issn
1868-4238
-
dc.description.startpage
349
-
dc.description.endpage
361
-
dc.type.category
Full-Paper Contribution
-
dc.relation.eissn
1868-422X
-
dc.publisher.place
Greece
-
tuw.booktitle
Artificial Intelligence Applications and Innovations - 17th IFIP WG 12.5 International Conference, AIAI 2021, Hersonissos, Crete, Greece, June 25–27, 2021, Proceedings
-
tuw.container.volume
627
-
tuw.peerreviewed
true
-
tuw.relation.publisher
Springer
-
tuw.relation.publisherplace
Cham
-
tuw.researchTopic.id
I4a
-
tuw.researchTopic.id
I2
-
tuw.researchTopic.name
Information Systems Engineering
-
tuw.researchTopic.name
Computer Engineering and Software-Intensive Systems