publications.bib

@misc{taketsugu_active_2023,
	title = {Active {Transfer} {Learning} for {Efficient} {Video}-{Specific} {Human} {Pose} {Estimation}},
	copyright = {All rights reserved},
	url = {http://arxiv.org/abs/2311.05041},
	doi = {10.48550/arXiv.2311.05041},
	abstract = {Human Pose (HP) estimation is actively researched because of its wide range of applications. However, even estimators pre-trained on large datasets may not perform satisfactorily due to a domain gap between the training and test data. To address this issue, we present our approach combining Active Learning (AL) and Transfer Learning (TL) to adapt HP estimators to individual video domains efficiently. For efficient learning, our approach quantifies (i) the estimation uncertainty based on the temporal changes in the estimated heatmaps and (ii) the unnaturalness in the estimated full-body HPs. These quantified criteria are then effectively combined with the state-of-the-art representativeness criterion to select uncertain and diverse samples for efficient HP estimator learning. Furthermore, we reconsider the existing Active Transfer Learning (ATL) method to introduce novel ideas related to the retraining methods and Stopping Criteria (SC). Experimental results demonstrate that our method enhances learning efficiency and outperforms comparative methods. Our code is publicly available at: https://github.com/ImIntheMiddle/VATL4Pose-WACV2024},
	urldate = {2023-11-24},
	publisher = {arXiv},
	author = {Taketsugu, Hiromu and Ukita, Norimichi},
	month = nov,
	year = {2023},
	note = {arXiv:2311.05041 [cs]},
	annote = {Comment: 17 pages, 12 figures, Accepted by WACV 2024},
}

@inproceedings{taketsugu_uncertainty_2023,
	title = {Uncertainty {Criteria} in {Active} {Transfer} {Learning} for {Efficient} {Video}-{Specific} {Human} {Pose} {Estimation}},
	copyright = {All rights reserved},
	url = {https://ieeexplore.ieee.org/abstract/document/10215565},
	doi = {10.23919/MVA57639.2023.10215565},
	abstract = {This paper presents a combination of Active Learning (AL) and Transfer Learning (TL) for efficiently adapting Human Pose (HP) estimators to individual videos. The proposed approach quantifies estimation uncertainty through the temporal changes and unnaturalness of estimated HPs. These uncertainty criteria are combined with clustering-based representativeness criterion to avoid the useless selection of similar samples. Experiments demonstrated that the proposed method achieves high learning efficiency and outperforms comparative methods.},
	urldate = {2023-11-24},
	booktitle = {2023 18th {International} {Conference} on {Machine} {Vision} and {Applications} ({MVA})},
	author = {Taketsugu, Hiromu and Ukita, Norimichi},
	month = jul,
	year = {2023},
	pages = {1--5},
}

@article{__2023,
	title = {効率的な動画別の人体姿勢推定を目的とした能動転移学習の不確実性基準},
	volume = {2023},
	url = {https://jglobal.jst.go.jp/detail?JGLOBAL_ID=202302214664049283},
	language = {ja},
	number = {CVIM-234},
	urldate = {2023-11-24},
	journal = {情報処理学会研究報告(Web)},
	author = {武次広夢 and 浮田宗伯},
	year = {2023},
	pages = {2023--234},
}