Accelerating Multi-Object Tracking in Edge Computing Environment with Time-Spatial Optimization
Multi-Object Tracking (MOT) task is a key issue in computer vision. It has capacious application prospects in survelliance, self-driving, and augmented reality. Benefit from the continuous progress of Deep Neural Network (DNN), the accuracy of the MOT algorithm has been significantly improved. Never...
Saved in:
Published in | 2021 Ninth International Conference on Advanced Cloud and Big Data (CBD) pp. 279 - 284 |
---|---|
Main Authors | , , , , , , , , |
Format | Conference Proceeding |
Language | English |
Published |
IEEE
01.03.2022
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Abstract | Multi-Object Tracking (MOT) task is a key issue in computer vision. It has capacious application prospects in survelliance, self-driving, and augmented reality. Benefit from the continuous progress of Deep Neural Network (DNN), the accuracy of the MOT algorithm has been significantly improved. Nevertheless, limited to computing power, achieving real-time DNN-based MOT is difficult in embedded systems. In reality, there are many wasteful and unnecessary computations in traditional frame-by-frame full-size video analysis. Therefore, in this paper, we propose a strategy that optimizing the execution of a traditional MOT pipeline in the dimension of time and space. In the temporal dimension, DNN only works in periodic keyframes while using a lightweight model for quickly generating results in the common frames. In the spatial dimension, we design an image density region discriminator to narrow down the input size of DNN. An edge device is introduced to perform end-edge collaborative computing to further accelerating the execution. Additionally, an end-edge parallel computing mechanism is designed that performing dynamic decisions based on the computing power and network environment between end and edge. Moreover, we rebuild the DNN model by TensorRT to optimize the model structure of DNN. By integrating the above approaches, the system can achieve 17.6 ~ 38.1 × speedup ratio, while with 3%~10.4% absolute tracking accuracy sacrifice and can be deployed in an unstable network environment. |
---|---|
AbstractList | Multi-Object Tracking (MOT) task is a key issue in computer vision. It has capacious application prospects in survelliance, self-driving, and augmented reality. Benefit from the continuous progress of Deep Neural Network (DNN), the accuracy of the MOT algorithm has been significantly improved. Nevertheless, limited to computing power, achieving real-time DNN-based MOT is difficult in embedded systems. In reality, there are many wasteful and unnecessary computations in traditional frame-by-frame full-size video analysis. Therefore, in this paper, we propose a strategy that optimizing the execution of a traditional MOT pipeline in the dimension of time and space. In the temporal dimension, DNN only works in periodic keyframes while using a lightweight model for quickly generating results in the common frames. In the spatial dimension, we design an image density region discriminator to narrow down the input size of DNN. An edge device is introduced to perform end-edge collaborative computing to further accelerating the execution. Additionally, an end-edge parallel computing mechanism is designed that performing dynamic decisions based on the computing power and network environment between end and edge. Moreover, we rebuild the DNN model by TensorRT to optimize the model structure of DNN. By integrating the above approaches, the system can achieve 17.6 ~ 38.1 × speedup ratio, while with 3%~10.4% absolute tracking accuracy sacrifice and can be deployed in an unstable network environment. |
Author | Liu, Mengyang Yin, Daheng Zhao, Wei Wang, Huitian Cai, Guangxing Dong, Fang Tang, Anran Shen, Lin Chang, Yunhan |
Author_xml | – sequence: 1 givenname: Mengyang surname: Liu fullname: Liu, Mengyang email: myliu@seu.edu.cn organization: Southeast University,School of Computer Science and Engineering,Nanjing,China – sequence: 2 givenname: Anran surname: Tang fullname: Tang, Anran email: anrant@seu.edu.cn organization: Southeast University,School of Computer Science and Engineering,Nanjing,China – sequence: 3 givenname: Huitian surname: Wang fullname: Wang, Huitian email: huitwang@seu.edu.cn organization: Southeast University,School of Computer Science and Engineering,Nanjing,China – sequence: 4 givenname: Lin surname: Shen fullname: Shen, Lin email: linshen@seu.edu.cn organization: Southeast University,School of Computer Science and Engineering,Nanjing,China – sequence: 5 givenname: Yunhan surname: Chang fullname: Chang, Yunhan email: zachchang@seu.edu.cn organization: Southeast University,School of Computer Science and Engineering,Nanjing,China – sequence: 6 givenname: Guangxing surname: Cai fullname: Cai, Guangxing email: guangxingcai@seu.edu.cn organization: Southeast University,School of Computer Science and Engineering,Nanjing,China – sequence: 7 givenname: Daheng surname: Yin fullname: Yin, Daheng email: yindaheng98@seu.edu.cn organization: Southeast University,School of Computer Science and Engineering,Nanjing,China – sequence: 8 givenname: Fang surname: Dong fullname: Dong, Fang email: fdong@seu.edu.cn organization: Southeast University,School of Computer Science and Engineering,Nanjing,China – sequence: 9 givenname: Wei surname: Zhao fullname: Zhao, Wei email: zhaoweistuart@gmail.com organization: Comprehensive National Science Center,Institute for Artificial Intelligence,Hefei,China |
BookMark | eNotjMtOwzAUBY0EC1r4gm78Awm-iX2dLEsoFKkoC7JgFzl-FEPiRMEFwddTHqvRGR3NgpyGMVhCVsBSAFZeVdc3giPINGMZpIwxIU7IAhAFZ5KLp3PSrrW2vZ1V9GFPHw599EndvVgdaTMr_fpjfaAbs7e0Gofp8PvbhHc_j2GwIdIPH59p4webPE7HiuppPUU_-K_jGMMFOXOqf7OX_1yS5nbTVNtkV9_dV-td4nlRJlJyrTW4DArMQXYuNwzRKDAgjHGaOdU5lAI7VqIrDHa8NJqXwMBwCZgvyeov66217TT7Qc2fbVkAZgLyby4cUas |
CODEN | IEEPAD |
ContentType | Conference Proceeding |
DBID | 6IE 6IL CBEJK RIE RIL |
DOI | 10.1109/CBD54617.2021.00055 |
DatabaseName | IEEE Electronic Library (IEL) Conference Proceedings IEEE Proceedings Order Plan All Online (POP All Online) 1998-present by volume IEEE Xplore All Conference Proceedings IEEE/IET Electronic Library (IEL) IEEE Proceedings Order Plans (POP All) 1998-Present |
DatabaseTitleList | |
Database_xml | – sequence: 1 dbid: RIE name: IEEE/IET Electronic Library (IEL) url: https://proxy.k.utb.cz/login?url=https://ieeexplore.ieee.org/ sourceTypes: Publisher |
DeliveryMethod | fulltext_linktorsrc |
EISBN | 166540745X 9781665407458 |
EndPage | 284 |
ExternalDocumentID | 9816251 |
Genre | orig-research |
GrantInformation_xml | – fundername: National Natural Science Foundation of China funderid: 10.13039/501100001809 – fundername: Ministry of Education funderid: 10.13039/100010002 – fundername: Natural Science Foundation of Jiangsu Province funderid: 10.13039/501100004608 |
GroupedDBID | 6IE 6IL CBEJK RIE RIL |
ID | FETCH-LOGICAL-i489-774ccc1f2186317bf3d066da1d15ddfc0fabf6756b096f8d6b49dc49101d47163 |
IEDL.DBID | RIE |
IngestDate | Thu Jun 29 18:37:09 EDT 2023 |
IsPeerReviewed | false |
IsScholarly | false |
Language | English |
LinkModel | DirectLink |
MergedId | FETCHMERGED-LOGICAL-i489-774ccc1f2186317bf3d066da1d15ddfc0fabf6756b096f8d6b49dc49101d47163 |
PageCount | 6 |
ParticipantIDs | ieee_primary_9816251 |
PublicationCentury | 2000 |
PublicationDate | 2022-March |
PublicationDateYYYYMMDD | 2022-03-01 |
PublicationDate_xml | – month: 03 year: 2022 text: 2022-March |
PublicationDecade | 2020 |
PublicationTitle | 2021 Ninth International Conference on Advanced Cloud and Big Data (CBD) |
PublicationTitleAbbrev | CBD |
PublicationYear | 2022 |
Publisher | IEEE |
Publisher_xml | – name: IEEE |
Score | 1.8411329 |
Snippet | Multi-Object Tracking (MOT) task is a key issue in computer vision. It has capacious application prospects in survelliance, self-driving, and augmented... |
SourceID | ieee |
SourceType | Publisher |
StartPage | 279 |
SubjectTerms | Bandwidth Collaboration Computational modeling Deep Learning Edge Intelligence Image edge detection Inference Optimization Multi-Object Tracking Performance evaluation Pipelines Streaming media |
Title | Accelerating Multi-Object Tracking in Edge Computing Environment with Time-Spatial Optimization |
URI | https://ieeexplore.ieee.org/document/9816251 |
hasFullText | 1 |
inHoldings | 1 |
isFullTextHit | |
isPrint | |
link | http://utb.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwjV09T8MwELVKJyZALeJbHhhJGztOao9QWlVIpQxF6lbVZ7uqEAGhduHXc-eEFiEGtshLonOcd_dy7x1j1wBOpj4UidKGqBs8c6aXFQmA9LlLjdTR7Hn8WIye1cMsnzXYzVYL472PzWe-Q5fxX757gw1RZV2jBabrWOvs6VRWWq3aSEikptu_u88VAjIWfVKQLyHJ936MTImIMTxg4-97VY0iL53N2nbg85cN438f5pC1d9o8_rRFnSPW8GWLzW8BED9oN8slj6LaZGKJYuEIRkB0OF-VfOCWnldzHGhlsBO5ceJjOelBEhpSjC8ln-DH5LVWabbZdDiY9kdJPTohWWHQMWVWACACDZzCBMGGzGFq4RbCidy5AGlY2IClQmGxggnaFVYZBwpTB-EQrYrsmDXLt9KfMJ5bkCKkmc98TxnQurBWeDKlN4FkPaesRbGZv1fmGPM6LGd_L5-zfUn6gdjEdcGa64-Nv0RUX9uruJ1fVYikTQ |
link.rule.ids | 310,311,786,790,795,796,802,27956,55107 |
linkProvider | IEEE |
linkToHtml | http://utb.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwjV09T8MwELWqMsAEqEV844GRtHG-PUJpVaAfDEXqFtVnu6oQKULtwq_nzgktQgxskZdE5zjv7uXeO8auAXTgG5t4USaJusEzJ9Mw8QACE2tfBpkzex6Okv5L9DiNpzV2s9HCGGNc85lp0aX7l6-XsCaqrC0zgek61jo7iPN-Wqq1Kish4ct25-4-jhCSsewLBDkTkoDvx9AUhxm9fTb8vlvZKvLaWq9UCz5_GTH-93EOWHOrzuPPG9w5ZDVTNFh-C4AIQvtZzLmT1XpjRSQLRzgCIsT5ouBdPTe8nORAK92tzI0TI8tJEeLRmGJ8LfkYPydvlU6zySa97qTT96rhCd4Cw45JcwQAwtLIKUwRlA01Jhd6JrSItbbg25myWCwkCmsYm-lERVJDhMmD0IhXSXjE6sWyMMeMxwoCYf3QhCaNJGRZopQwZEsvLQl7TliDYpO_l_YYeRWW07-Xr9hufzIc5IOH0dMZ2wtITeBaus5ZffWxNheI8St16bb2CyrCp6E |
openUrl | ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Abook&rft.genre=proceeding&rft.title=2021+Ninth+International+Conference+on+Advanced+Cloud+and+Big+Data+%28CBD%29&rft.atitle=Accelerating+Multi-Object+Tracking+in+Edge+Computing+Environment+with+Time-Spatial+Optimization&rft.au=Liu%2C+Mengyang&rft.au=Tang%2C+Anran&rft.au=Wang%2C+Huitian&rft.au=Shen%2C+Lin&rft.date=2022-03-01&rft.pub=IEEE&rft.spage=279&rft.epage=284&rft_id=info:doi/10.1109%2FCBD54617.2021.00055&rft.externalDocID=9816251 |