Resource-constrained implementation and optimization of a deep neural network for vehicle classification

Xie, Renjie; Huttunen, Heikki; Lin, Shiming; Bhattacharyya, Shuvra S.; Takala, Jarmo

doi:10.1109/eusipco.2016.7760571

Cited by 12 publications

(7 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Research on teaching models has probably gone through three stages: introduction and practical precipitation in the 1980s, theoretical research and disciplinary construction in the 1990s, and practical construction and regional promotion in the 21st century. Although after decades of active exploration and practice, teaching models have been effectively developed and progressed [4], each teaching mode has its own advantages and shortcomings, and no teaching mode is perfect. erefore, a single teaching mode cannot meet the needs of complex teaching, and in specific teaching practice, we should move from a single mode construction to the comprehensive application of various modes, absorb the essence of various teaching modes, learn from the strengths of all, optimize the combination of modes, and pursue the optimization of teaching effect.…”

Section: Introductionmentioning

confidence: 99%

Classroom Resource Optimization of the English Four-Step Model Based on Deep Learning

Zhao

Wang

2021

Security and Communication Networks

View full text Add to dashboard Cite

The existing LBL (lecture-based learning) and CBL (case-based learning) teaching modes of English majors in colleges and universities do not contribute to cultivating students’ inquisitive thinking and independent learning ability, whereas the problem-centered PBL (problem-based learning) teaching mode can precisely make up for these shortcomings. This study firstly constructed a PBL “four-step” teaching mode and then realized the innovative application of PBL teaching model in the teaching of English majors in new universities. Then, we designed a method of English classroom resource optimization based on IMOCS-BP neural network, and then, the classroom resource optimization reconstructs the teaching process with unique connotation essence. The experiments show that the BL “four-step” teaching model is designed to achieve the four key elements of “flexible English learning environment, transformed learning the culture, customized content and professional educators”; the designed neural network has optimized the English classroom resources.

show abstract

Section: Introductionmentioning

confidence: 99%

Classroom Resource Optimization of the English Four-Step Model Based on Deep Learning

Zhao

Wang

2021

Security and Communication Networks

View full text Add to dashboard Cite

show abstract

“…3) Adaptive Deep Neural Network: The Deep Neural Network (DNN) application for vehicle classification has first been presented in [29], [30]. The neural network consists of two convolutional layers followed by three dense layers.…”

Section: Resultsmentioning

confidence: 99%

PRUNE: Dynamic and Decidable Dataflow for Signal Processing on Heterogeneous Platforms

Boutellier

Huttunen

et al. 2018

IEEE Trans. Signal Process.

Self Cite

View full text Add to dashboard Cite

Abstract-The majority of contemporary mobile devices and personal computers are based on heterogeneous computing platforms that consist of a number of CPU cores and one or more Graphics Processing Units (GPUs). Despite the high volume of these devices, there are few existing programming frameworks that target full and simultaneous utilization of all CPU and GPU devices of the platform.This article presents a dataflow-flavored Model of Computation (MoC) that has been developed for deploying signal processing applications to heterogeneous platforms. The presented MoC is dynamic and allows describing applications with data dependent run-time behavior. On top of the MoC, formal design rules are presented that enable application descriptions to be simultaneously dynamic and decidable. Decidability guarantees compile-time application analyzability for deadlock freedom and bounded memory.The presented MoC and the design rules are realized in a novel Open Source programming environment "PRUNE" and demonstrated with representative application examples from the domains of image processing, computer vision and wireless communications. Experimental results show that the proposed approach outperforms the state-of-the-art in analyzability, flexibility and performance.

show abstract

“…The experimental evaluation of Edge-PRUNE was conducted using two convolutional neural networks: an image classification network for vehicle image classification [28], and object tracking based on the SSD-Mobilenet object detector [26]. The experimental platforms (see Table I) cover an Intel Core i7 based system that acts as the edge server, and two lightweight systems that act as endpoint devices: an ARM multicore single board computer ODROID N2 with a Mali G-52 GPU, and an Intel Atom N270 based single-core system.…”

Section: Methodsmentioning

confidence: 99%

“…The CNN for vehicle image classification [28] consists of two convolutional layers with 5×5 filter size, max-pooling by a downsampling factor of two, and ReLU activation. The two convolutional layers are followed by three dense layers, of which the first two use ReLU activation, whereas the third dense layer is followed by SoftMax.…”

Section: A Convolutional Neural Network Use Casesmentioning

confidence: 99%

Edge-PRUNE: Flexible Distributed Deep Learning Inference

Boutellier¹,

Tan²,

Nurmi³

2022

Preprint

View full text Add to dashboard Cite

Collaborative deep learning inference between lowresource endpoint devices and edge servers has received significant research interest in the last few years. Such computation partitioning can help reducing endpoint device energy consumption and improve latency, but equally importantly also contributes to privacy-preserving of sensitive data. This paper describes Edge-PRUNE, a flexible but light-weight computation framework for distributing machine learning inference between edge servers and one or more client devices. Compared to previous approaches, Edge-PRUNE is based on a formal dataflow computing model, and is agnostic towards machine learning training frameworks, offering at the same time wide support for leveraging deep learning accelerators such as embedded GPUs. The experimental section of the paper demonstrates the use and performance of Edge-PRUNE by image classification and object tracking applications on two heterogeneous endpoint devices and an edge server, over wireless and physical connections. Endpoint device inference time for SSD-Mobilenet based object tracking, for example, is accelerated 5.8× by collaborative inference.

show abstract

Resource-constrained implementation and optimization of a deep neural network for vehicle classification

Cited by 12 publications

References 7 publications

Classroom Resource Optimization of the English Four-Step Model Based on Deep Learning

Classroom Resource Optimization of the English Four-Step Model Based on Deep Learning

PRUNE: Dynamic and Decidable Dataflow for Signal Processing on Heterogeneous Platforms

Edge-PRUNE: Flexible Distributed Deep Learning Inference

Contact Info

Product

Resources

About