DTTNet: Improving Video Shadow Detection via Dark-Aware Guidance and Tokenized Temporal Modeling
Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (8) · 6753-6761
DOI: 10.1609/aaai.v40i8.37607 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Zhicheng Li und 7 weiteren Autor:innen.
Abstract (Originalsprache)
Video shadow detection confronts two entwined difficulties: distinguishing shadows from complex backgrounds and modeling dynamic shadow deformations under varying illumination. To address shadow-background ambiguity, we leverage linguistic priors through the proposed Vision-language Match Module (VMM) and a Dark-aware Semantic Block (DSB), extracting text-guided features to explicitly differentiate shadows from dark objects. Furthermore, we introduce adaptive mask reweighting to downweight penumbra regions during training and apply edge masks at the final decoder stage for better supervision. For temporal modeling of variable shadow shapes, we propose a Tokenized Temporal Block (TTB) that decouples spatiotemporal learning. TTB summarizes cross-frame shadow semantics into learnable temporal tokens, enabling efficient sequence encoding with minimal computation overhead. Comprehensive Experiments on multiple benchmark datasets demonstrate state-of-the-art accuracy and real-time inference efficiency.
Zitieren
APA
Li, Z., Sun, K., Yao, R., Zhu, H., Hu, F., Zhao, J., Shao, Z., & Zhou, Y. (2026). DTTNet: Improving Video Shadow Detection via Dark-Aware Guidance and Tokenized Temporal Modeling. Proceedings of the AAAI Conference on Artificial Intelligence, 40(8), 6753-6761. https://doi.org/10.1609/aaai.v40i8.37607
Vancouver
Li Z, Sun K, Yao R, Zhu H, Hu F, Zhao J, et al. DTTNet: Improving Video Shadow Detection via Dark-Aware Guidance and Tokenized Temporal Modeling. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(8):6753-6761. doi:10.1609/aaai.v40i8.37607
BibTeX
@article{Li2026,
author = {Li, Zhicheng and Sun, Kunyang and Yao, Rui and Zhu, Hancheng and Hu, Fuyuan and Zhao, Jiaqi and Shao, Zhiwen and Zhou, Yong},
title = {DTTNet: Improving Video Shadow Detection via Dark-Aware Guidance and Tokenized Temporal Modeling},
journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2026},
volume = {40},
number = {8},
pages = {6753--6761},
doi = {10.1609/aaai.v40i8.37607}
}
EUVIMED