Show simple item record

dc.contributor.authorPad, Pedram
dc.contributor.authorNarduzzi, Simon
dc.contributor.authorKündig, Clément
dc.contributor.authorTüretken, Engin
dc.contributor.authorBigdeli, Siavash A.
dc.contributor.authorDunbar, L. Andrea
dc.date.accessioned2022-02-21T10:08:15Z
dc.date.available2022-02-21T10:08:15Z
dc.date.issued2020-06-14
dc.identifier.citationIn Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Seattle, WA, US, pp. 12285-12294en_US
dc.identifier.isbn978-1-7281-7168-5
dc.identifier.urihttps://yoda.csem.ch/handle/20.500.12839/994
dc.description.abstractDespite the substantial progress made in deep learning in recent years, advanced approaches remain computationally intensive. The trade-off between accuracy and computation time and energy limits their use in real-time applications on low power and other resource-constrained systems. In this paper, we tackle this fundamental challenge by introducing a hybrid optical-digital implementation of a convolutional neural network (CNN) based on engineering of the point spread function (PSF) of an optical imaging system. This is done by coding an imaging aperture such that its PSF replicates a large convolution kernel of the first layer of a pre-trained CNN. As the convolution takes place in the optical domain, it has zero cost in terms of energy consumption and has zero latency independent of the kernel size. Experimental results on two datasets demonstrate that our approach yields more than two orders of magnitude reduction in the computational cost while achieving near-state-of-the-art accuracy, or equivalently, better accuracy at the same computational cost.en_US
dc.language.isoenen_US
dc.subjectEmbedded Deep Learningen_US
dc.subjectOpticsen_US
dc.subjectKernelen_US
dc.subjectErgoen_US
dc.titleEfficient Neural Vision Systems Based on Convolutional Image Acquisitionen_US
dc.typeProceedingsen_US
dc.type.csemdivisionsDiv-Men_US
dc.type.csemresearchareasData & AIen_US
dc.type.csemresearchareasASICs for the Edgeen_US
dc.type.csemresearchareasIoT & Visionen_US


Files in this item

Thumbnail

This item appears in the following Collection(s)

  • Research Publications
    The “Research Publications” collection provides bibliographic information for scientific papers including conference proceedings and presentations.

Show simple item record