CVNet: Lightweight Cross-View Vehicle ReID with Multi-Scale Localization

Cross-view vehicle re-identification (ReID) between aerial and ground perspectives is challenging due to limited computational resources on edge devices and significant scale variations. We propose CVNet, a lightweight network with two key modules: the multi-scale localization (MSL) module and the d...

Full description

Saved in:
Bibliographic Details
Main Authors: Wenji Yin, Baixuan Han, Yueping Peng, Hexiang Hao, Zecong Ye, Yu Shen, Yanjun Cai, Wenchao Kang
Format: Article
Language:English
Published: MDPI AG 2025-04-01
Series:Sensors
Subjects:
Online Access:https://www.mdpi.com/1424-8220/25/9/2809
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850135702696624128
author Wenji Yin
Baixuan Han
Yueping Peng
Hexiang Hao
Zecong Ye
Yu Shen
Yanjun Cai
Wenchao Kang
author_facet Wenji Yin
Baixuan Han
Yueping Peng
Hexiang Hao
Zecong Ye
Yu Shen
Yanjun Cai
Wenchao Kang
author_sort Wenji Yin
collection DOAJ
description Cross-view vehicle re-identification (ReID) between aerial and ground perspectives is challenging due to limited computational resources on edge devices and significant scale variations. We propose CVNet, a lightweight network with two key modules: the multi-scale localization (MSL) module and the deep–shallow filtrate collaboration (DFC) module. The MSL module employs multi-scale depthwise separable convolutions and a localization attention mechanism to extract multi-scale features and localize salient regions, addressing viewpoint variations. DFC employs a dual-branch design comprising deep and shallow branches, integrating a filtration module optimized via neural architecture search, a collaboration module, and lightweight convolutions. This design effectively captures both unique and shared cross-view features, ensuring efficient and robust feature representation. We also release a new CVPair v1.0 dataset, the first benchmark for cross-view ReID, containing 14,969 images of 894 vehicle identities, offering results of traditional and lightweight methods. CVNet achieves state-of-the-art performance on CVPair v1.0, VehicleID, and VeRi776, advancing cross-view vehicle ReID. The dataset will be released publicly.
format Article
id doaj-art-ab081ea696cd46588dda20bead7df37f
institution OA Journals
issn 1424-8220
language English
publishDate 2025-04-01
publisher MDPI AG
record_format Article
series Sensors
spelling doaj-art-ab081ea696cd46588dda20bead7df37f2025-08-20T02:31:20ZengMDPI AGSensors1424-82202025-04-01259280910.3390/s25092809CVNet: Lightweight Cross-View Vehicle ReID with Multi-Scale LocalizationWenji Yin0Baixuan Han1Yueping Peng2Hexiang Hao3Zecong Ye4Yu Shen5Yanjun Cai6Wenchao Kang7School of Information Engineering, PAP Engineering University, Xi’an 710086, ChinaSchool of Information Engineering, PAP Engineering University, Xi’an 710086, ChinaSchool of Information Engineering, PAP Engineering University, Xi’an 710086, ChinaSchool of Information Engineering, PAP Engineering University, Xi’an 710086, ChinaSchool of Information Engineering, PAP Engineering University, Xi’an 710086, ChinaSchool of Information Engineering, PAP Engineering University, Xi’an 710086, ChinaSchool of Information Engineering, PAP Engineering University, Xi’an 710086, ChinaSchool of Information Engineering, PAP Engineering University, Xi’an 710086, ChinaCross-view vehicle re-identification (ReID) between aerial and ground perspectives is challenging due to limited computational resources on edge devices and significant scale variations. We propose CVNet, a lightweight network with two key modules: the multi-scale localization (MSL) module and the deep–shallow filtrate collaboration (DFC) module. The MSL module employs multi-scale depthwise separable convolutions and a localization attention mechanism to extract multi-scale features and localize salient regions, addressing viewpoint variations. DFC employs a dual-branch design comprising deep and shallow branches, integrating a filtration module optimized via neural architecture search, a collaboration module, and lightweight convolutions. This design effectively captures both unique and shared cross-view features, ensuring efficient and robust feature representation. We also release a new CVPair v1.0 dataset, the first benchmark for cross-view ReID, containing 14,969 images of 894 vehicle identities, offering results of traditional and lightweight methods. CVNet achieves state-of-the-art performance on CVPair v1.0, VehicleID, and VeRi776, advancing cross-view vehicle ReID. The dataset will be released publicly.https://www.mdpi.com/1424-8220/25/9/2809re-identificationcross-viewlightweight network
spellingShingle Wenji Yin
Baixuan Han
Yueping Peng
Hexiang Hao
Zecong Ye
Yu Shen
Yanjun Cai
Wenchao Kang
CVNet: Lightweight Cross-View Vehicle ReID with Multi-Scale Localization
Sensors
re-identification
cross-view
lightweight network
title CVNet: Lightweight Cross-View Vehicle ReID with Multi-Scale Localization
title_full CVNet: Lightweight Cross-View Vehicle ReID with Multi-Scale Localization
title_fullStr CVNet: Lightweight Cross-View Vehicle ReID with Multi-Scale Localization
title_full_unstemmed CVNet: Lightweight Cross-View Vehicle ReID with Multi-Scale Localization
title_short CVNet: Lightweight Cross-View Vehicle ReID with Multi-Scale Localization
title_sort cvnet lightweight cross view vehicle reid with multi scale localization
topic re-identification
cross-view
lightweight network
url https://www.mdpi.com/1424-8220/25/9/2809
work_keys_str_mv AT wenjiyin cvnetlightweightcrossviewvehiclereidwithmultiscalelocalization
AT baixuanhan cvnetlightweightcrossviewvehiclereidwithmultiscalelocalization
AT yuepingpeng cvnetlightweightcrossviewvehiclereidwithmultiscalelocalization
AT hexianghao cvnetlightweightcrossviewvehiclereidwithmultiscalelocalization
AT zecongye cvnetlightweightcrossviewvehiclereidwithmultiscalelocalization
AT yushen cvnetlightweightcrossviewvehiclereidwithmultiscalelocalization
AT yanjuncai cvnetlightweightcrossviewvehiclereidwithmultiscalelocalization
AT wenchaokang cvnetlightweightcrossviewvehiclereidwithmultiscalelocalization