Optimization of Lightweight Convolutional Neural Networks for Pedestrian Recognition

Kaifu Dai

doi:10.47297/taposatWSP2633-456911.20230402

PDF
Export
Share
Collection
Album

Optimization of Lightweight Convolutional Neural Networks for Pedestrian Recognition

Volume 4, Issue 2

Views: 0 Downloads: 0

Optimization of Lightweight Convolutional Neural Networks for Pedestrian Recognition
School of Computer and Information Technology, Hohai University,Jiangsu,Nanjing,P.R.China,211100
Vol. 4, Issue 2, Pages: 73-77(2023)
Published： 2023 ，
DOI：10.47297/taposatWSP2633-456911.20230402

Full txt

Kaifu Dai. (2023). Optimization of Lightweight Convolutional Neural Networks for Pedestrian Recognition. Theory and Practice of Science and Technology, 4(2), 73-77.
DOI：

Kaifu Dai. (2023). Optimization of Lightweight Convolutional Neural Networks for Pedestrian Recognition. Theory and Practice of Science and Technology, 4(2), 73-77. DOI： 10.47297/taposatWSP2633-456911.20230402.

摘要

Abstract

Based on the wide application of pedestrian recognition in the fields of smart city and automatic driving

a suitable deep neural network model is selected for the devices that are in embedded systems and with poor computing power. In this paper

by establishing a pedestrian behavior data set

the MobileNet series depth separable convolution structure is optimized and improved. MobileNetV3 is pruned and optimized in the course of the experiment

and suitable hyper parameters are selected. The experimental results show that under the condition of training with Nadam optimizer

when the width multiplier coefficient α is 0.6 and the resolution multiplier coefficient ρ is 0.9

the results of the MobileNetV3 Small model are optimal and the accuracy can reach 94.4%.This model has fewer parameters and computational effort than the traditional neural network model

and is more suitable for deployment on embedded systems and devices with poor computing power.

关键词

Keywords

LightweightPedestrian recognitionDeep neural network

references

Andrew G. Howard, Menglong Zhu, Bo Chen, Dmitry Kalenichenko, Weijun Wang, Tobias Weyand, Marco Andreetto, Hartwig Adam. (2017) MobileNets:Efficient Convolutional Neural Networks for Mobile Vision Applications. https://www.researchgate.net/publication/316184205_MobileNets_Efficient_Convolutional_Neural_Networks_for_Mobile_Vision_Applications[2] Lu Haiyan, Zhao Hongdong, Wang Tianmeng, etc. (2022) Infrared pedestrian behavior recognition based on lightweight convolutional neural network. Sensors and Microsystems, 2022, 41(09): 129-31. [3] Du F J, Jiao S J. Improvement of lightweight convolutional neural network model based on YOLO algorithm and its research in pavement defect detection[J]. Sensors, 2022, 22(9): 3537.[4] Nupoor Yawale. (2023) Design of a high-density bio-inspired feature analysis deep learning model for sub-classification of natural & synthetic imagery. https://www.researchgate.net/publication/372940300_Design_of_a_high-density_bio-inspired_feature_analysis_deep_learning_model_for_sub-classification_of_natural_synthetic_imagery?_sg=SRsbV9r_LO93c5UnPvxnqKHzoSABS3Axvq7nvY55OBHGFwszuBJR5rHV-NCj8cr_IvjSmPlJ0oRvYMM [5] HERMOSILLA P,RITSCHEL T,VAZQUEZ P. (2018) Monte Carlo convolution for learning on non-uniformly sampled point clouds. ACM Transactions on Graphics(TOG),2018,37(6):1-12.[6] Wang J, Li H, Yin S, et al. Research on improved pedestrian detection algorithm based on convolutional neural network[C]//2019 international conference on Internet of Things (iThings) and IEEE green computing and communications (GreenCom) and IEEE cyber, physical and social computing (CPSCom) and IEEE smart data (SmartData). IEEE, 2019: 254-58.

Alert me when the article has been cited

Submit

No data

Related Author

No data

Related Institution

No data

⁰