https://www.mdu.se/

mdu.sePublications
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
DenseDisp: Resource-Aware Disparity Map Estimation by Compressing Siamese Neural Architecture
Mälardalen University, School of Innovation, Design and Engineering, Embedded Systems.ORCID iD: 0000-0002-9704-7117
Shiraz Univ Technol, Shiraz, Iran.
Technische Universität Berlin, Germany.
Abo Akad Univ, Dept Informat Technol, Turku, Finland.
Show others and affiliations
2020 (English)In: IEEE WORLD CONGRESS ON COMPUTATIONAL INTELLIGENCE (WCCI) 2020 IEEE WCCI, Glasgow, United Kingdom, 2020Conference paper, Published paper (Refereed)
Abstract [en]

Stereo vision cameras are flexible sensors due to providing heterogeneous information such as color, luminance, disparity map (depth), and shape of the objects. Today, Convolutional Neural Networks (CNNs) present the highest accuracy for the disparity map estimation [1]. However, CNNs require considerable computing capacity to process billions of floating-point operations in a real-time fashion. Besides, commercial stereo cameras produce huge size images (e.g., 10 Megapixels [2]), which impose a new computational cost to the system. The problem will be pronounced if we target resource-limited hardware for the implementation. In this paper, we propose DenseDisp, an automatic framework that designs a Siamese neural architecture for disparity map estimation in a reasonable time. DenseDisp leverages a meta-heuristic multi-objective exploration to discover hardware-friendly architectures by considering accuracy and network FLOPS as the optimization objectives. We explore the design space with four different fitness functions to improve the accuracy-FLOPS trade-off and convergency time of the DenseDisp. According to the experimental results, DenseDisp provides up to 39.1x compression rate while losing around 5% accuracy compared to the state-of-the-art results.

Place, publisher, year, edition, pages
Glasgow, United Kingdom, 2020.
Keywords [en]
Stereo Vision, Deep Learning, Multi-Objective, Optimization, Neural Architecture Search
National Category
Engineering and Technology Computer Systems
Identifiers
URN: urn:nbn:se:mdh:diva-49331ISI: 000703998201002Scopus ID: 2-s2.0-85055448153OAI: oai:DiVA.org:mdh-49331DiVA, id: diva2:1453356
Conference
IEEE WORLD CONGRESS ON COMPUTATIONAL INTELLIGENCE (WCCI) 2020 IEEE WCCI, 19 Jul 2020, Glasgow, United Kingdom
Projects
DeepMaker: Deep Learning Accelerator on Commercial Programmable DevicesAvailable from: 2020-07-09 Created: 2020-07-09 Last updated: 2023-05-10Bibliographically approved
In thesis
1. DeepMaker: Customizing the Architecture of Convolutional Neural Networks for Resource-Constrained Platforms
Open this publication in new window or tab >>DeepMaker: Customizing the Architecture of Convolutional Neural Networks for Resource-Constrained Platforms
2020 (English)Licentiate thesis, comprehensive summary (Other academic)
Abstract [en]

Convolutional Neural Networks (CNNs) suffer from energy-hungry implementation due to requiring huge amounts of computations and significant memory consumption. This problem will be more highlighted by the proliferation of CNNs on resource-constrained platforms in, e.g., embedded systems. In this thesis, we focus on decreasing the computational cost of CNNs in order to be appropriate for resource-constrained platforms. The thesis work proposes two distinct methods to tackle the challenges: optimizing CNN architecture while considering network accuracy and network complexity, and proposing an optimized ternary neural network to compensate the accuracy loss of network quantization methods. We evaluated the impact of our solutions on Commercial-Off-The-Shelf (COTS) platforms where the results show considerable improvement in network accuracy and energy efficiency.

Abstract [sv]

Convolutional Neural Networks (CNNs) lider av energihungriga implementationer på grund av att de kräver enorm beräkningskapacitet och har en betydande minneskonsumtion. Detta problem kommer att framhävas mer när allt fler CNN implementeras på resursbegränsade plattformar i inbyggda datorsystem. I denna uppsats fokuserar vi på att minska resursåtgången för CNN, i termer av behövda beräkningar och behövt minne, för att vara lämplig för resursbegränsade plattformar. Vi föreslår två metoder för att hantera utmaningarna; optimera CNN-arkitektur där man balanserar nätverksnoggrannhet och nätverkskomplexitet, och föreslår ett optimerat ternärt neuralt nätverk för att kompensera noggrannhetsförluster som kan uppstå vid nätverkskvantiseringsmetoder. Vi utvärderade effekterna av våra lösningar på kommersiellt använda plattformar (COTS) där resultaten visar betydande förbättringar i nätverksnoggrannhet och energieffektivitet.

Place, publisher, year, edition, pages
Västerås: Mälardalen University, 2020
Series
Mälardalen University Press Licentiate Theses, ISSN 1651-9256 ; 299
National Category
Computer Sciences
Research subject
Computer Science
Identifiers
urn:nbn:se:mdh:diva-52113 (URN)978-91-7485-490-9 (ISBN)
Presentation
2020-12-04, U2-024 (+ Online/Zoom), Mälardalens högskola, Västerås, 11:30 (English)
Opponent
Supervisors
Projects
DeepMaker: Deep Learning Accelerator on Commercial Programmable DevicesDPAC - Dependable Platforms for Autonomous systems and ControlFAST-ARTS: Fast and Sustainable Analysis Techniques for Advanced Real-Time Systems
Available from: 2020-11-10 Created: 2020-10-29 Last updated: 2020-11-13Bibliographically approved

Open Access in DiVA

No full text in DiVA

Scopus

Authority records

Loni, MohammadDaneshtalab, MasoudSjödin, Mikael

Search in DiVA

By author/editor
Loni, MohammadDaneshtalab, MasoudSjödin, Mikael
By organisation
Embedded Systems
Engineering and TechnologyComputer Systems

Search outside of DiVA

GoogleGoogle Scholar

urn-nbn

Altmetric score

urn-nbn
Total: 121 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf