Please use this identifier to cite or link to this item:
https://doi.org/10.1109/TCYB.2021.3130047
DC Field | Value | |
---|---|---|
dc.title | Soft Person Reidentification Network Pruning via Blockwise Adjacent Filter Decaying | |
dc.contributor.author | Wang, Xiaodong | |
dc.contributor.author | Zheng, Zhedong | |
dc.contributor.author | He, Yang | |
dc.contributor.author | Yan, Fei | |
dc.contributor.author | Zeng, Zhiqiang | |
dc.contributor.author | Yang, Yi | |
dc.date.accessioned | 2023-11-14T03:17:12Z | |
dc.date.available | 2023-11-14T03:17:12Z | |
dc.date.issued | 2022-12 | |
dc.identifier.citation | Wang, Xiaodong, Zheng, Zhedong, He, Yang, Yan, Fei, Zeng, Zhiqiang, Yang, Yi (2022-12). Soft Person Reidentification Network Pruning via Blockwise Adjacent Filter Decaying. IEEE TRANSACTIONS ON CYBERNETICS 52 (12) : 13293-13307. ScholarBank@NUS Repository. https://doi.org/10.1109/TCYB.2021.3130047 | |
dc.identifier.issn | 2168-2267 | |
dc.identifier.issn | 2168-2275 | |
dc.identifier.uri | https://scholarbank.nus.edu.sg/handle/10635/245918 | |
dc.description.abstract | Deep learning has shown significant successes in person reidentification (re-id) tasks. However, most existing works focus on discriminative feature learning and impose complex neural networks, suffering from low inference efficiency. In fact, feature extraction time is also crucial for real-world applications and lightweight models are needed. Prevailing pruning methods usually pay attention to compact classification models. However, these methods are suboptimal for compacting re-id models, which usually produce continuous features and are sensitive to network pruning. The key point of pruning re-id models is how to retain the original filter distribution in continuous features as much as possible. In this work, we propose a blockwise adjacent filter decaying method to fill this gap. Specifically, given a trained model, we first evaluate the redundancy of filters based on the adjacency relationships to preserve the original filter distribution. Second, previous layerwise pruning methods ignore that discriminative information is enhanced block-by-block. Therefore, we propose a blockwise filter pruning strategy to better utilize the block relations in the pretrained model. Third, we propose a novel filter decaying policy to progressively reduce the scale of redundant filters. Different from conventional soft filter pruning that directly sets the filter values as zeros, the proposed filter decaying can keep the pretrained knowledge as much as possible. We evaluate our method on three popular person reidentification datasets, that is: 1) Market-1501; 2) DukeMTMC-reID; and 3) MSMT17_V1. The proposed method shows superior performance to the existing state-of-the-art pruning methods. After pruning over 91.9% parameters on DukeMTMC-reID, the Rank-1 accuracy only drops 3.7%, demonstrating its effectiveness for compacting person reidentification. | |
dc.language.iso | en | |
dc.publisher | IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC | |
dc.source | Elements | |
dc.subject | Science & Technology | |
dc.subject | Technology | |
dc.subject | Automation & Control Systems | |
dc.subject | Computer Science, Artificial Intelligence | |
dc.subject | Computer Science, Cybernetics | |
dc.subject | Computer Science | |
dc.subject | Redundancy | |
dc.subject | Feature extraction | |
dc.subject | Training | |
dc.subject | Neural networks | |
dc.subject | Convolutional neural networks | |
dc.subject | Computational modeling | |
dc.subject | Topology | |
dc.subject | Deep learning | |
dc.subject | network pruning | |
dc.subject | person reidentification | |
dc.subject | representation learning | |
dc.type | Article | |
dc.date.updated | 2023-11-11T03:34:37Z | |
dc.contributor.department | DEPARTMENT OF COMPUTER SCIENCE | |
dc.contributor.department | ELECTRICAL AND COMPUTER ENGINEERING | |
dc.description.doi | 10.1109/TCYB.2021.3130047 | |
dc.description.sourcetitle | IEEE TRANSACTIONS ON CYBERNETICS | |
dc.description.volume | 52 | |
dc.description.issue | 12 | |
dc.description.page | 13293-13307 | |
dc.published.state | Published | |
Appears in Collections: | Elements Staff Publications |
Show simple item record
Files in This Item:
File | Description | Size | Format | Access Settings | Version | |
---|---|---|---|---|---|---|
Wang_Soft.pdf | Accepted version | 2.89 MB | Adobe PDF | OPEN | Post-print | View/Download |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.