Tree-based Unidirectional Neural Networks for Low-Power Computer Vision

Abhinav Goel; Caleb Tung; Nick Eliopoulos; Amy Wang; Jamie C Davis; George K. Thiruvathukal; Yung-Hisang Lu

doi:10.1109/MDAT.2022.3217016

Tree-based Unidirectional Neural Networks for Low-Power Computer Vision

Abhinav Goel, Caleb Tung, Nick Eliopoulos, Amy Wang, Jamie C Davis, George K. Thiruvathukal, Yung-Hisang Lu

Department of Computer Science

Research output: Contribution to journal › Article › peer-review

Abstract

This article describes the novel Tree-based Unidirectional Neural Network (TRUNK) architecture. This architecture improves computer vision efficiency by using a hierarchy of multiple shallow Convolutional Neural Networks (CNNs), instead of a single very deep CNN. We demonstrate this architecture’s versatility in performing different computer vision tasks efficiently on embedded devices. Across various computer vision tasks, the TRUNK architecture consumes 65% less energy and requires 50% less memory than representative low-power CNN architectures, e.g., MobileNet v2, when deployed on the NVIDIA Jetson Nano.

Original language	American English
Journal	Computer Science: Faculty Publications and Other Works
Volume	40
Issue number	3
DOIs	https://doi.org/10.1109/MDAT.2022.3217016
State	Published - Jun 1 2023

Keywords

Measurement
Convolutional neural networks
Visualization
Computer vision
Task analysis
Memory management
Horses

Disciplines

Artificial Intelligence and Robotics
Computer Sciences

Access to Document

10.1109/MDAT.2022.3217016License: Unspecified

Tree-based Unidirectional Neural Networks for Low-Power ComputerFinal published version, 2.08 MBLicense: Unspecified

Cite this

@article{a8a6025eba0349ada2214e4f0a6b6a34,

title = "Tree-based Unidirectional Neural Networks for Low-Power Computer Vision",

abstract = " This article describes the novel Tree-based Unidirectional Neural Network (TRUNK) architecture. This architecture improves computer vision efficiency by using a hierarchy of multiple shallow Convolutional Neural Networks (CNNs), instead of a single very deep CNN. We demonstrate this architecture{\textquoteright}s versatility in performing different computer vision tasks efficiently on embedded devices. Across various computer vision tasks, the TRUNK architecture consumes 65% less energy and requires 50% less memory than representative low-power CNN architectures, e.g., MobileNet v2, when deployed on the NVIDIA Jetson Nano.",

keywords = "Measurement, Convolutional neural networks, Visualization, Computer vision, Task analysis, Memory management, Horses",

author = "Abhinav Goel and Caleb Tung and Nick Eliopoulos and Amy Wang and Davis, {Jamie C} and Thiruvathukal, {George K.} and Yung-Hisang Lu",

note = "A. Goel, C. Tung, N. Eliopoulos, A. Wang, J.C. Davis, G.K. Thiruvathukal, and Y.-H. Lu, {"}Tree-based Unidirectional Neural Networks for Low-Power Computer Vision,{"} in IEEE Design & Test, 2022, doi: 10.1109/MDAT.2022.3217016.",

year = "2023",

month = jun,

day = "1",

doi = "10.1109/MDAT.2022.3217016",

language = "American English",

volume = "40",

journal = "Computer Science: Faculty Publications and Other Works",

number = "3",

}

TY - JOUR

T1 - Tree-based Unidirectional Neural Networks for Low-Power Computer Vision

AU - Goel, Abhinav

AU - Tung, Caleb

AU - Eliopoulos, Nick

AU - Wang, Amy

AU - Davis, Jamie C

AU - Thiruvathukal, George K.

AU - Lu, Yung-Hisang

N1 - A. Goel, C. Tung, N. Eliopoulos, A. Wang, J.C. Davis, G.K. Thiruvathukal, and Y.-H. Lu, "Tree-based Unidirectional Neural Networks for Low-Power Computer Vision," in IEEE Design & Test, 2022, doi: 10.1109/MDAT.2022.3217016.

PY - 2023/6/1

Y1 - 2023/6/1

N2 - This article describes the novel Tree-based Unidirectional Neural Network (TRUNK) architecture. This architecture improves computer vision efficiency by using a hierarchy of multiple shallow Convolutional Neural Networks (CNNs), instead of a single very deep CNN. We demonstrate this architecture’s versatility in performing different computer vision tasks efficiently on embedded devices. Across various computer vision tasks, the TRUNK architecture consumes 65% less energy and requires 50% less memory than representative low-power CNN architectures, e.g., MobileNet v2, when deployed on the NVIDIA Jetson Nano.

AB - This article describes the novel Tree-based Unidirectional Neural Network (TRUNK) architecture. This architecture improves computer vision efficiency by using a hierarchy of multiple shallow Convolutional Neural Networks (CNNs), instead of a single very deep CNN. We demonstrate this architecture’s versatility in performing different computer vision tasks efficiently on embedded devices. Across various computer vision tasks, the TRUNK architecture consumes 65% less energy and requires 50% less memory than representative low-power CNN architectures, e.g., MobileNet v2, when deployed on the NVIDIA Jetson Nano.

KW - Measurement

KW - Convolutional neural networks

KW - Visualization

KW - Computer vision

KW - Task analysis

KW - Memory management

KW - Horses

UR - https://ecommons.luc.edu/cs_facpubs/317

U2 - 10.1109/MDAT.2022.3217016

DO - 10.1109/MDAT.2022.3217016

M3 - Article

VL - 40

JO - Computer Science: Faculty Publications and Other Works

JF - Computer Science: Faculty Publications and Other Works

IS - 3

ER -

Tree-based Unidirectional Neural Networks for Low-Power Computer Vision

Abstract

Keywords

Disciplines

Access to Document

Other files and links

Cite this