Split Computing With Scalable Feature Compression for Visual Analytics on the Edge

被引:0
作者
Yuan, Zhongzheng [1 ]
Rawlekar, Samyak [2 ]
Garg, Siddharth [1 ]
Erkip, Elza [1 ]
Wang, Yao [1 ]
机构
[1] NYU, Elect & Comp Engn Dept, Tandon Sch Engn, Brooklyn, NY 11201 USA
[2] Univ Illinois, Dept Elect & Comp Engn, Urbana, IL 61801 USA
关键词
Image coding; Computational modeling; Task analysis; Servers; Performance evaluation; Bit rate; Analytical models; Computer vision; feature compression; object detection; split computing; COLLABORATIVE INTELLIGENCE;
D O I
10.1109/TMM.2024.3406165
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Running deep visual analytics models for real-time applications is challenging for mobile devices. Offloading the computation to edge server can mitigate computation bottleneck at the mobile device, but may decrease the analytics performance due to the necessity of compressing the image data. We consider a "split computing" system to offload a part of the deep learning model's computation and introduce a novel learned feature compression approach with lightweight computation. We demonstrate the effectiveness of the split computing pipeline in performing computation offloading for the problems of object detection and image classification. Compared to compressing the raw images at the mobile, and running the analytics model on the decompressed images at the server, the proposed feature-compression approach can achieve significantly higher analytics performance at the same bit rate, while reducing the complexity at the mobile. We further propose a scalable feature compression approach, which facilitates adaptation to network bandwidth dynamics, while having comparable performance to the non-scalable approach.
引用
收藏
页码:10121 / 10133
页数:13
相关论文
共 50 条
[41]   Intelligent Resource Allocation for Video Analytics in Blockchain-Enabled Internet of Autonomous Vehicles With Edge Computing [J].
Jiang, Xiantao ;
Yu, F. Richard ;
Song, Tian ;
Leung, Victor C. M. .
IEEE INTERNET OF THINGS JOURNAL, 2022, 9 (16) :14260-14272
[42]   Leveraging Spatial Attention and Edge Context for Optimized Feature Selection in Visual Localization [J].
Istighfarin, Nanda Febri ;
Jo, Hyunggi .
INTERNATIONAL JOURNAL OF CONTROL AUTOMATION AND SYSTEMS, 2025, 23 (02) :418-428
[43]   Edge Devices Clustering for Federated Visual Classification: A Feature Norm Based Framework [J].
Wei, Xiao-Xiang ;
Huang, Hua .
IEEE TRANSACTIONS ON IMAGE PROCESSING, 2023, 32 :995-1010
[44]   Attention-based variable-size feature compression module for edge inference [J].
Shibao Li ;
Chenxu Ma ;
Yunwu Zhang ;
Longfei Li ;
Chengzhi Wang ;
Xuerong Cui ;
Jianhang Liu .
The Journal of Supercomputing, 2024, 80 :8469-8484
[45]   Attention-based variable-size feature compression module for edge inference [J].
Li, Shibao ;
Ma, Chenxu ;
Zhang, Yunwu ;
Li, Longfei ;
Wang, Chengzhi ;
Cui, Xuerong ;
Liu, Jianhang .
JOURNAL OF SUPERCOMPUTING, 2024, 80 (06) :8469-8484
[46]   Smart Network Maintenance in an Edge Cloud Computing Environment: An Adaptive Model Compression Algorithm Based on Model Pruning and Model Clustering [J].
Rui, Lanlan ;
Yang, Siqi ;
Chen, Shiyou ;
Yang, Yang ;
Gao, Zhipeng .
IEEE TRANSACTIONS ON NETWORK AND SERVICE MANAGEMENT, 2022, 19 (04) :4165-4175
[47]   Joint Split Offloading and Trajectory Scheduling for UAV-Enabled Mobile Edge Computing in IoT Network [J].
Wei, Yunkai ;
Wan, Zikang ;
Xiao, Yinan ;
Leng, Supeng ;
Wang, Kezhi ;
Yang, Kun .
IEEE TRANSACTIONS ON NETWORK SCIENCE AND ENGINEERING, 2024, 11 (06) :6180-6193
[48]   Dynamic Split Computing-Aware Mixed-Precision Quantization for Efficient Deep Edge Intelligence [J].
Nagamatsu, Naoki ;
Hara-Azumi, Yuko .
2023 IEEE 22ND INTERNATIONAL CONFERENCE ON TRUST, SECURITY AND PRIVACY IN COMPUTING AND COMMUNICATIONS, TRUSTCOM, BIGDATASE, CSE, EUC, ISCI 2023, 2024, :2538-2545
[49]   LO-SC: Local-only Split Computing for Accurate Deep Learning on Edge Devices [J].
Capogrosso, Luigi ;
Fraccaroli, Enrico ;
Cristani, Marco ;
Fummi, Franco ;
Chakraborty, Samarjit .
2025 38TH INTERNATIONAL CONFERENCE ON VLSI DESIGN AND 2024 23RD INTERNATIONAL CONFERENCE ON EMBEDDED SYSTEMS, VLSID, 2025, :445-450
[50]   Trustworthy and Scalable Federated Edge Learning for Future Integrated Positioning, Communication, and Computing System: Attacks and Defenses [J].
Zhang, Ronghui ;
Wu, Sheng ;
Jiang, Chunxiao ;
Gao, Ning ;
Qiu, Xuesong ;
Zhang, Wei .
IEEE INTERNET OF THINGS JOURNAL, 2024, 11 (21) :34243-34253