DTTNet: Improving Video Shadow Detection via Dark-Aware Guidance and Tokenized Temporal Modeling

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Li, Zhicheng, Sun, Kunyang, Yao, Rui, Zhu, Hancheng, Hu, Fuyuan, Zhao, Jiaqi, Shao, Zhiwen, Zhou, Yong
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909895950860288
author Li, Zhicheng
Sun, Kunyang
Yao, Rui
Zhu, Hancheng
Hu, Fuyuan
Zhao, Jiaqi
Shao, Zhiwen
Zhou, Yong
author_facet Li, Zhicheng
Sun, Kunyang
Yao, Rui
Zhu, Hancheng
Hu, Fuyuan
Zhao, Jiaqi
Shao, Zhiwen
Zhou, Yong
contents 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. Codes are available at https://github.com/city-cheng/DTTNet.
format Preprint
id arxiv_https___arxiv_org_abs_2511_06925
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle DTTNet: Improving Video Shadow Detection via Dark-Aware Guidance and Tokenized Temporal Modeling
Li, Zhicheng
Sun, Kunyang
Yao, Rui
Zhu, Hancheng
Hu, Fuyuan
Zhao, Jiaqi
Shao, Zhiwen
Zhou, Yong
Computer Vision and Pattern Recognition
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. Codes are available at https://github.com/city-cheng/DTTNet.
title DTTNet: Improving Video Shadow Detection via Dark-Aware Guidance and Tokenized Temporal Modeling
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2511.06925