Pseudo label refining for semi-supervised temporal action localization.

The training of temporal action localization models relies heavily on a large amount of manually annotated data. Video annotation is more tedious and time-consuming compared with image annotation. Therefore, the semi-supervised method that combines labeled and unlabeled data for joint training has a...

Full description

Saved in:
Bibliographic Details
Main Authors: Lingwen Meng, Guobang Ban, Guanghui Xi, Siqi Guo
Format: Article
Language:English
Published: Public Library of Science (PLoS) 2025-01-01
Series:PLoS ONE
Online Access:https://doi.org/10.1371/journal.pone.0318418
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850187763632046080
author Lingwen Meng
Guobang Ban
Guanghui Xi
Siqi Guo
author_facet Lingwen Meng
Guobang Ban
Guanghui Xi
Siqi Guo
author_sort Lingwen Meng
collection DOAJ
description The training of temporal action localization models relies heavily on a large amount of manually annotated data. Video annotation is more tedious and time-consuming compared with image annotation. Therefore, the semi-supervised method that combines labeled and unlabeled data for joint training has attracted increasing attention from academics and industry. This study proposes a method called pseudo-label refining (PLR) based on the teacher-student framework, which consists of three key components. First, we propose pseudo-label self-refinement which features in a temporal region interesting pooling to improve the boundary accuracy of TAL pseudo label. Second, we design a module named boundary synthesis to further refined temporal interval in pseudo label with multiple inference. Finally, an adaptive weight learning strategy is tailored for progressively learning pseudo labels with different qualities. The method proposed in this study uses ActionFormer and BMN as the detector and achieves significant improvement on the THUMOS14 and ActivityNet v1.3 datasets. The experimental results show that the proposed method significantly improve the localization accuracy compared to other advanced SSTAL methods at a label rate of 10% to 60%. Further ablation experiments show the effectiveness of each module, proving that the PLR method can improve the accuracy of pseudo-labels obtained by teacher model reasoning.
format Article
id doaj-art-0ad7ef4ced3b4c8fa69740da9926aebd
institution OA Journals
issn 1932-6203
language English
publishDate 2025-01-01
publisher Public Library of Science (PLoS)
record_format Article
series PLoS ONE
spelling doaj-art-0ad7ef4ced3b4c8fa69740da9926aebd2025-08-20T02:16:02ZengPublic Library of Science (PLoS)PLoS ONE1932-62032025-01-01202e031841810.1371/journal.pone.0318418Pseudo label refining for semi-supervised temporal action localization.Lingwen MengGuobang BanGuanghui XiSiqi GuoThe training of temporal action localization models relies heavily on a large amount of manually annotated data. Video annotation is more tedious and time-consuming compared with image annotation. Therefore, the semi-supervised method that combines labeled and unlabeled data for joint training has attracted increasing attention from academics and industry. This study proposes a method called pseudo-label refining (PLR) based on the teacher-student framework, which consists of three key components. First, we propose pseudo-label self-refinement which features in a temporal region interesting pooling to improve the boundary accuracy of TAL pseudo label. Second, we design a module named boundary synthesis to further refined temporal interval in pseudo label with multiple inference. Finally, an adaptive weight learning strategy is tailored for progressively learning pseudo labels with different qualities. The method proposed in this study uses ActionFormer and BMN as the detector and achieves significant improvement on the THUMOS14 and ActivityNet v1.3 datasets. The experimental results show that the proposed method significantly improve the localization accuracy compared to other advanced SSTAL methods at a label rate of 10% to 60%. Further ablation experiments show the effectiveness of each module, proving that the PLR method can improve the accuracy of pseudo-labels obtained by teacher model reasoning.https://doi.org/10.1371/journal.pone.0318418
spellingShingle Lingwen Meng
Guobang Ban
Guanghui Xi
Siqi Guo
Pseudo label refining for semi-supervised temporal action localization.
PLoS ONE
title Pseudo label refining for semi-supervised temporal action localization.
title_full Pseudo label refining for semi-supervised temporal action localization.
title_fullStr Pseudo label refining for semi-supervised temporal action localization.
title_full_unstemmed Pseudo label refining for semi-supervised temporal action localization.
title_short Pseudo label refining for semi-supervised temporal action localization.
title_sort pseudo label refining for semi supervised temporal action localization
url https://doi.org/10.1371/journal.pone.0318418
work_keys_str_mv AT lingwenmeng pseudolabelrefiningforsemisupervisedtemporalactionlocalization
AT guobangban pseudolabelrefiningforsemisupervisedtemporalactionlocalization
AT guanghuixi pseudolabelrefiningforsemisupervisedtemporalactionlocalization
AT siqiguo pseudolabelrefiningforsemisupervisedtemporalactionlocalization