dc.creator | Aimar, Alessandro | es |
dc.creator | Mostafa, Hesham | es |
dc.creator | Calabrese, Enrico | es |
dc.creator | Ríos Navarro, José Antonio | es |
dc.creator | Tapiador Morales, Ricardo | es |
dc.creator | Lungu, Iulia-Alexandra | es |
dc.creator | Milde, Moritz B. | es |
dc.creator | Corradi, Federico | es |
dc.creator | Linares Barranco, Alejandro | es |
dc.creator | Liu, Shih-Chii | es |
dc.creator | Delbruck, Tobi | es |
dc.date.accessioned | 2020-01-31T11:54:40Z | |
dc.date.available | 2020-01-31T11:54:40Z | |
dc.date.issued | 2019 | |
dc.identifier.citation | Aimar, A., Mostafa, H., Calabrese, E., Ríos Navarro, J.A., Tapiador Morales, R., Lungu, I.,...,Delbruck, T. (2019). NullHop: A Flexible Convolutional Neural Network Accelerator Based on Sparse Representations of Feature Maps. IEEE Transactions on Neural Networks and Learning Systems, 30 (3), 644-656. | |
dc.identifier.issn | 2162-237X | es |
dc.identifier.uri | https://hdl.handle.net/11441/92660 | |
dc.description.abstract | Convolutional neural networks (CNNs) have become
the dominant neural network architecture for solving many stateof-
the-art (SOA) visual processing tasks. Even though Graphical
Processing Units (GPUs) are most often used in training and deploying
CNNs, their power efficiency is less than 10 GOp/s/W for
single-frame runtime inference.We propose a flexible and efficient
CNN accelerator architecture called NullHop that implements
SOA CNNs useful for low-power and low-latency application
scenarios. NullHop exploits the sparsity of neuron activations
in CNNs to accelerate the computation and reduce memory
requirements. The flexible architecture allows high utilization of
available computing resources across kernel sizes ranging from
1x1 to 7x7. NullHop can process up to 128 input and 128 output
feature maps per layer in a single pass. We implemented the
proposed architecture on a Xilinx Zynq FPGA platform and
present results showing how our implementation reduces external
memory transfers and compute time in five different CNNs
ranging from small ones up to the widely known large VGG16 and
VGG19 CNNs. Post-synthesis simulations using Mentor Modelsim
in a 28nm process with a clock frequency of 500MHz show
that the VGG19 network achieves over 450GOp/s. By exploiting
sparsity, NullHop achieves an efficiency of 368%, maintains over
98% utilization of the MAC units, and achieves a power efficiency
of over 3TOp/s/W in a core area of 6.3mm2. As further proof
of NullHop’s usability, we interfaced its FPGA implementation
with a neuromorphic event camera for real time interactive
demonstrations. | es |
dc.format | application/pdf | es |
dc.language.iso | eng | es |
dc.publisher | IEEE Computer Society | es |
dc.relation.ispartof | IEEE Transactions on Neural Networks and Learning Systems, 30 (3), 644-656. | |
dc.rights | Attribution-NonCommercial-NoDerivatives 4.0 Internacional | * |
dc.rights.uri | http://creativecommons.org/licenses/by-nc-nd/4.0/ | * |
dc.subject | Convolutional Neural Networks (CNN) | es |
dc.subject | VLSI | es |
dc.subject | FPGA | es |
dc.subject | Computer vision | es |
dc.subject | Artificial intelligence | es |
dc.title | NullHop: A Flexible Convolutional Neural Network Accelerator Based on Sparse Representations of Feature Maps | es |
dc.type | info:eu-repo/semantics/article | es |
dcterms.identifier | https://ror.org/03yxnpp24 | |
dc.type.version | info:eu-repo/semantics/submittedVersion | es |
dc.rights.accessRights | info:eu-repo/semantics/openAccess | es |
dc.contributor.affiliation | Universidad de Sevilla. Departamento de Arquitectura y Tecnología de Computadores | es |
dc.relation.publisherversion | https://ieeexplore.ieee.org/document/8421093 | es |
dc.identifier.doi | 10.1109/TNNLS.2018.2852335 | es |
dc.contributor.group | Universidad de Sevilla. TEP-108: Robótica y Tecnología de Computadores Aplicada a la Rehabilitación | es |
idus.format.extent | 13 | es |
dc.journaltitle | IEEE Transactions on Neural Networks and Learning Systems | es |
dc.publication.volumen | 30 | es |
dc.publication.issue | 3 | es |
dc.publication.initialPage | 644 | es |
dc.publication.endPage | 656 | es |