SS‐SoS: Sample Selected Based Salvage of Supervision Framework for Weakly Supervised Object Detection
IET Computer Vision · 2026-01 · 20 (1)
DOI: 10.1049/cvi2.70084 ↗Kurz gefasst
Fachartikel in der Zeitschrift IET Computer Vision, online erschienen am 09.11.2026. Verfasst von Yihuan Zhu und 3 weiteren Autor:innen. Die Arbeit nennt 81 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0). Gefördert von: National Natural Science Foundation of China.
Abstract (Originalsprache)
Weakly Supervised Object Detection (WSOD) suffers from a notable performance gap compared to Fully Supervised Object Detection (FSOD), as it fails to exploit localisation cues and modern detection advances. To bridge this gap, we propose a novel framework named Sample Selected Based Salvage of Supervision (SS‐SoS). This framework adopts a progressive pipeline from weakly supervised object detection to semi‐supervised object detection (SSOD) to fully supervised object detection, where the dependency on localisation supervision increases gradually across stages. To ensure robust training, SS‐SoS employs a classification‐localisation consistency based WSOD model and introduces an Uncertainty‐based Multi‐round Training (UMT) sample selection strategy. This strategy effectively separates clean and noisy samples whilst preserving hard examples through dynamic modelling. Extensive experiments on VOC 2007, VOC 2012 and MS‐COCO validate the effectiveness of our method on WSOD tasks. Further evaluations on CUB‐200 and ILSVRC confirm its superiority in weakly supervised object localisation (WSOL), achieving state‐of‐the‐art performance across all benchmarks.
Förderung
- National Natural Science Foundation of China
Zitieren
APA
Zhu, Y., Wang, S., Lu, M., & Sun, Z. (2026). SS‐SoS: Sample Selected Based Salvage of Supervision Framework for Weakly Supervised Object Detection. IET Computer Vision, 20(1). https://doi.org/10.1049/cvi2.70084
Vancouver
Zhu Y, Wang S, Lu M, Sun Z. SS‐SoS: Sample Selected Based Salvage of Supervision Framework for Weakly Supervised Object Detection. IET Computer Vision. 2026;20(1). doi:10.1049/cvi2.70084
BibTeX
@article{Zhu2026,
author = {Zhu, Yihuan and Wang, Simiao and Lu, Mingyu and Sun, Zhengxing},
title = {SS‐SoS: Sample Selected Based Salvage of Supervision Framework for Weakly Supervised Object Detection},
journal = {IET Computer Vision},
year = {2026},
volume = {20},
number = {1},
doi = {10.1049/cvi2.70084}
}
EUVIMED