Noise-resistant Deep Metric Learning with Ranking-based Instance Selection
The existence of noisy labels in real-world data negatively impacts the performance of deep learning models. Although much research effort has been devoted to improving robustness to noisy labels in classification tasks, the problem of noisy labels in deep metric learning (DML) remains open. In this...
Main Authors: | , , , , , , , , |
---|---|
Format: | Article in Journal/Newspaper |
Language: | unknown |
Published: |
arXiv
2021
|
Subjects: | |
Online Access: | https://dx.doi.org/10.48550/arxiv.2103.16047 https://arxiv.org/abs/2103.16047 |
id |
ftdatacite:10.48550/arxiv.2103.16047 |
---|---|
record_format |
openpolar |
spelling |
ftdatacite:10.48550/arxiv.2103.16047 2023-05-15T16:01:35+02:00 Noise-resistant Deep Metric Learning with Ranking-based Instance Selection Liu, Chang Yu, Han Li, Boyang Shen, Zhiqi Gao, Zhanning Ren, Peiran Xie, Xuansong Cui, Lizhen Miao, Chunyan 2021 https://dx.doi.org/10.48550/arxiv.2103.16047 https://arxiv.org/abs/2103.16047 unknown arXiv Creative Commons Attribution 4.0 International https://creativecommons.org/licenses/by/4.0/legalcode cc-by-4.0 CC-BY Computer Vision and Pattern Recognition cs.CV FOS Computer and information sciences Article CreativeWork article Preprint 2021 ftdatacite https://doi.org/10.48550/arxiv.2103.16047 2022-03-10T14:48:51Z The existence of noisy labels in real-world data negatively impacts the performance of deep learning models. Although much research effort has been devoted to improving robustness to noisy labels in classification tasks, the problem of noisy labels in deep metric learning (DML) remains open. In this paper, we propose a noise-resistant training technique for DML, which we name Probabilistic Ranking-based Instance Selection with Memory (PRISM). PRISM identifies noisy data in a minibatch using average similarity against image features extracted by several previous versions of the neural network. These features are stored in and retrieved from a memory bank. To alleviate the high computational cost brought by the memory bank, we introduce an acceleration method that replaces individual data points with the class centers. In extensive comparisons with 12 existing approaches under both synthetic and real-world label noise, PRISM demonstrates superior performance of up to 6.06% in Precision@1. : Accepted by CVPR 2021 Article in Journal/Newspaper DML DataCite Metadata Store (German National Library of Science and Technology) |
institution |
Open Polar |
collection |
DataCite Metadata Store (German National Library of Science and Technology) |
op_collection_id |
ftdatacite |
language |
unknown |
topic |
Computer Vision and Pattern Recognition cs.CV FOS Computer and information sciences |
spellingShingle |
Computer Vision and Pattern Recognition cs.CV FOS Computer and information sciences Liu, Chang Yu, Han Li, Boyang Shen, Zhiqi Gao, Zhanning Ren, Peiran Xie, Xuansong Cui, Lizhen Miao, Chunyan Noise-resistant Deep Metric Learning with Ranking-based Instance Selection |
topic_facet |
Computer Vision and Pattern Recognition cs.CV FOS Computer and information sciences |
description |
The existence of noisy labels in real-world data negatively impacts the performance of deep learning models. Although much research effort has been devoted to improving robustness to noisy labels in classification tasks, the problem of noisy labels in deep metric learning (DML) remains open. In this paper, we propose a noise-resistant training technique for DML, which we name Probabilistic Ranking-based Instance Selection with Memory (PRISM). PRISM identifies noisy data in a minibatch using average similarity against image features extracted by several previous versions of the neural network. These features are stored in and retrieved from a memory bank. To alleviate the high computational cost brought by the memory bank, we introduce an acceleration method that replaces individual data points with the class centers. In extensive comparisons with 12 existing approaches under both synthetic and real-world label noise, PRISM demonstrates superior performance of up to 6.06% in Precision@1. : Accepted by CVPR 2021 |
format |
Article in Journal/Newspaper |
author |
Liu, Chang Yu, Han Li, Boyang Shen, Zhiqi Gao, Zhanning Ren, Peiran Xie, Xuansong Cui, Lizhen Miao, Chunyan |
author_facet |
Liu, Chang Yu, Han Li, Boyang Shen, Zhiqi Gao, Zhanning Ren, Peiran Xie, Xuansong Cui, Lizhen Miao, Chunyan |
author_sort |
Liu, Chang |
title |
Noise-resistant Deep Metric Learning with Ranking-based Instance Selection |
title_short |
Noise-resistant Deep Metric Learning with Ranking-based Instance Selection |
title_full |
Noise-resistant Deep Metric Learning with Ranking-based Instance Selection |
title_fullStr |
Noise-resistant Deep Metric Learning with Ranking-based Instance Selection |
title_full_unstemmed |
Noise-resistant Deep Metric Learning with Ranking-based Instance Selection |
title_sort |
noise-resistant deep metric learning with ranking-based instance selection |
publisher |
arXiv |
publishDate |
2021 |
url |
https://dx.doi.org/10.48550/arxiv.2103.16047 https://arxiv.org/abs/2103.16047 |
genre |
DML |
genre_facet |
DML |
op_rights |
Creative Commons Attribution 4.0 International https://creativecommons.org/licenses/by/4.0/legalcode cc-by-4.0 |
op_rightsnorm |
CC-BY |
op_doi |
https://doi.org/10.48550/arxiv.2103.16047 |
_version_ |
1766397373139910656 |