Show simple item record

dc.contributor.authorFornt Mas, Jordi
dc.contributor.authorFontova Muste, Pau
dc.contributor.authorCaro Roca, Martí
dc.contributor.authorAbella Ferrer, Jaume
dc.contributor.authorMoll Echeto, Francisco de Borja
dc.contributor.authorAltet Sanahujes, Josep
dc.contributor.authorRubio Sola, Jose Antonio
dc.contributor.otherUniversitat Politècnica de Catalunya. Doctorat en Enginyeria Electrònica
dc.contributor.otherUniversitat Politècnica de Catalunya. Doctorat en Arquitectura de Computadors
dc.contributor.otherUniversitat Politècnica de Catalunya. Departament d'Enginyeria Electrònica
dc.contributor.otherBarcelona Supercomputing Center
dc.date.accessioned2024-01-18T09:09:09Z
dc.date.available2024-01-18T09:09:09Z
dc.date.issued2023
dc.identifier.citationFornt, J. [et al.]. Energy efficient object detection for automotive applications with YOLOv3 and approximate hardware. A: IEEE International Conference on Nanotechnology. "2023 IEEE 23rd International Conference on Nanotechnology (IEEE-NANO 2023): Jeju, Korea: July 2-5, 2023: proceedings". Institute of Electrical and Electronics Engineers (IEEE), 2023, p. 592-595. ISBN 979-8-3503-3347-3. DOI 10.1109/NANO58406.2023.10231293.
dc.identifier.isbn979-8-3503-3347-3
dc.identifier.urihttp://hdl.handle.net/2117/399770
dc.description.abstractDeep neural networks are the dominant models for perception tasks in the automotive domain, but their high computational complexity makes it difficult to execute them in real time with an acceptable power consumption on general-purpose devices. For this reason, the design of custom ASIC devices for real-time energy-efficient neural network deployment is a hot topic in academia and industry. Low-precision integer arithmetic and approximate computing are two popular optimizations often contemplated for saving hardware resources and power consumption. In this work, we evaluate two approximate computing circuits using different integer precisions in order to study the trade-offs that these techniques offer in terms of energy efficiency and degradation of the neural network outputs. In particular, we apply the approximations to the YOLOv3 object detection network, a popular model for critical applications in the automotive domain. By combining approximate arithmetic circuits with low precision we are able to reduce the power consumption of a MAD unit by over 50% compared to using only quantization.
dc.format.extent4 p.
dc.language.isoeng
dc.publisherInstitute of Electrical and Electronics Engineers (IEEE)
dc.subjectÀrees temàtiques de la UPC::Enginyeria electrònica::Microelectrònica
dc.subjectÀrees temàtiques de la UPC::Enginyeria de la telecomunicació::Processament del senyal::Reconeixement de formes
dc.subject.lcshNeural networks (Computer science)
dc.subject.lcshPattern recognition systems
dc.subject.lcshEnergy consumption
dc.subject.otherApproximate computing
dc.subject.otherObject detection
dc.subject.otherLow-precision
dc.subject.otherQuantization
dc.subject.otherEnergy-efficient DNN acceleration
dc.titleEnergy efficient object detection for automotive applications with YOLOv3 and approximate hardware
dc.typeConference report
dc.subject.lemacXarxes neuronals (Informàtica)
dc.subject.lemacReconeixement de formes (Informàtica)
dc.subject.lemacEnergia -- Consum
dc.contributor.groupUniversitat Politècnica de Catalunya. EFRICS - Efficient and Robust Integrated Circuits and Systems
dc.identifier.doi10.1109/NANO58406.2023.10231293
dc.description.peerreviewedPeer Reviewed
dc.relation.publisherversionhttps://ieeexplore.ieee.org/document/10231293
dc.rights.accessOpen Access
local.identifier.drac37149585
dc.description.versionPostprint (author's final draft)
local.citation.authorFornt, J.; Fontova, P.; Caro, M.; Abella, J.; Moll, F.; Altet, J.; Rubio, A.
local.citation.contributorIEEE International Conference on Nanotechnology
local.citation.publicationName2023 IEEE 23rd International Conference on Nanotechnology (IEEE-NANO 2023): Jeju, Korea: July 2-5, 2023: proceedings
local.citation.startingPage592
local.citation.endingPage595


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record