DSC: Dense-sparse convolution for vectorized inference of convolutional neural networks

Alexander Frickenstein; Manoj Rohit Vemparala; Christian Unger; Fatih Ayar; Walter Stechele

doi:10.1109/CVPRW.2019.00175

DSC: Dense-sparse convolution for vectorized inference of convolutional neural networks

Alexander Frickenstein, Manoj Rohit Vemparala, Christian Unger, Fatih Ayar, Walter Stechele

Default

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

14 Scopus citations

Abstract

The efficient applications of Convolutional Neural Networks (CNNs) in automotive-rated and safety critical hardware-accelerators require an interplay of DNN design optimization, programming techniques and hardware resources. Ad-hoc pruning would result in irregular sparsity and compression leading in very inefficient real world applications. Therefore, the proposed methodology, called Dense-Sparse Convolution, makes use of the right balance between pruning regularity, quantization and the underlying vectorized hardware. Different word length compute units, e.g. CPU, are used for low latency inference of the spares CNNs. The proposed open source CPU-kernel scales along with the vector word length and the number of cores.

Original language	English
Title of host publication	Proceedings - 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2019
Publisher	IEEE Computer Society
Pages	1353-1360
Number of pages	8
ISBN (Electronic)	9781728125060
DOIs	https://doi.org/10.1109/CVPRW.2019.00175
State	Published - Jun 2019
Event	32nd IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2019 - Long Beach, United States Duration: 16 Jun 2019 → 20 Jun 2019

Publication series

Name	IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops
Volume	2019-June
ISSN (Print)	2160-7508
ISSN (Electronic)	2160-7516

Conference

Conference	32nd IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2019
Country/Territory	United States
City	Long Beach
Period	16/06/19 → 20/06/19

Access to Document

10.1109/CVPRW.2019.00175

Cite this

Frickenstein, A., Vemparala, M. R., Unger, C., Ayar, F., & Stechele, W. (2019). DSC: Dense-sparse convolution for vectorized inference of convolutional neural networks. In Proceedings - 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2019 (pp. 1353-1360). Article 9025437 (IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops; Vol. 2019-June). IEEE Computer Society. https://doi.org/10.1109/CVPRW.2019.00175

Frickenstein, Alexander ; Vemparala, Manoj Rohit ; Unger, Christian et al. / DSC : Dense-sparse convolution for vectorized inference of convolutional neural networks. Proceedings - 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2019. IEEE Computer Society, 2019. pp. 1353-1360 (IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops).

@inproceedings{292023343cad432a8d1c500fc395b99c,

title = "DSC: Dense-sparse convolution for vectorized inference of convolutional neural networks",

abstract = "The efficient applications of Convolutional Neural Networks (CNNs) in automotive-rated and safety critical hardware-accelerators require an interplay of DNN design optimization, programming techniques and hardware resources. Ad-hoc pruning would result in irregular sparsity and compression leading in very inefficient real world applications. Therefore, the proposed methodology, called Dense-Sparse Convolution, makes use of the right balance between pruning regularity, quantization and the underlying vectorized hardware. Different word length compute units, e.g. CPU, are used for low latency inference of the spares CNNs. The proposed open source CPU-kernel scales along with the vector word length and the number of cores.",

author = "Alexander Frickenstein and Vemparala, {Manoj Rohit} and Christian Unger and Fatih Ayar and Walter Stechele",

note = "Publisher Copyright: {\textcopyright} 2019 IEEE.; 32nd IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2019 ; Conference date: 16-06-2019 Through 20-06-2019",

year = "2019",

month = jun,

doi = "10.1109/CVPRW.2019.00175",

language = "English",

series = "IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops",

publisher = "IEEE Computer Society",

pages = "1353--1360",

booktitle = "Proceedings - 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2019",

}

Frickenstein, A, Vemparala, MR, Unger, C, Ayar, F & Stechele, W 2019, DSC: Dense-sparse convolution for vectorized inference of convolutional neural networks. in Proceedings - 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2019., 9025437, IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops, vol. 2019-June, IEEE Computer Society, pp. 1353-1360, 32nd IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2019, Long Beach, United States, 16/06/19. https://doi.org/10.1109/CVPRW.2019.00175

DSC: Dense-sparse convolution for vectorized inference of convolutional neural networks. / Frickenstein, Alexander; Vemparala, Manoj Rohit; Unger, Christian et al.
Proceedings - 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2019. IEEE Computer Society, 2019. p. 1353-1360 9025437 (IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops; Vol. 2019-June).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - DSC

T2 - 32nd IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2019

AU - Frickenstein, Alexander

AU - Vemparala, Manoj Rohit

AU - Unger, Christian

AU - Ayar, Fatih

AU - Stechele, Walter

PY - 2019/6

Y1 - 2019/6

N2 - The efficient applications of Convolutional Neural Networks (CNNs) in automotive-rated and safety critical hardware-accelerators require an interplay of DNN design optimization, programming techniques and hardware resources. Ad-hoc pruning would result in irregular sparsity and compression leading in very inefficient real world applications. Therefore, the proposed methodology, called Dense-Sparse Convolution, makes use of the right balance between pruning regularity, quantization and the underlying vectorized hardware. Different word length compute units, e.g. CPU, are used for low latency inference of the spares CNNs. The proposed open source CPU-kernel scales along with the vector word length and the number of cores.

AB - The efficient applications of Convolutional Neural Networks (CNNs) in automotive-rated and safety critical hardware-accelerators require an interplay of DNN design optimization, programming techniques and hardware resources. Ad-hoc pruning would result in irregular sparsity and compression leading in very inefficient real world applications. Therefore, the proposed methodology, called Dense-Sparse Convolution, makes use of the right balance between pruning regularity, quantization and the underlying vectorized hardware. Different word length compute units, e.g. CPU, are used for low latency inference of the spares CNNs. The proposed open source CPU-kernel scales along with the vector word length and the number of cores.

UR - http://www.scopus.com/inward/record.url?scp=85083341111&partnerID=8YFLogxK

U2 - 10.1109/CVPRW.2019.00175

DO - 10.1109/CVPRW.2019.00175

M3 - Conference contribution

AN - SCOPUS:85083341111

T3 - IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops

SP - 1353

EP - 1360

BT - Proceedings - 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2019

PB - IEEE Computer Society

Y2 - 16 June 2019 through 20 June 2019

ER -

Frickenstein A, Vemparala MR, Unger C, Ayar F, Stechele W. DSC: Dense-sparse convolution for vectorized inference of convolutional neural networks. In Proceedings - 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2019. IEEE Computer Society. 2019. p. 1353-1360. 9025437. (IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops). doi: 10.1109/CVPRW.2019.00175

DSC: Dense-sparse convolution for vectorized inference of convolutional neural networks

Abstract

Publication series

Conference

Access to Document

Other files and links

Fingerprint

Cite this