Physics-Grounded Attached Shadow Detection Using Approximate 3D Geometry and Light Direction

2025-12-05

AI Summary

Existing shadow detection methods predominantly focus on cast shadows while neglecting attached shadows—critical cues for 3D scene understanding—and lack dedicated datasets and models for them. To address this gap, we propose the first unified framework for joint detection of cast and attached shadows. Specifically, we introduce the first large-scale dataset annotated with attached shadows, comprising 1,458 images. We design a dual-branch network that jointly optimizes shadow detection and illumination estimation, incorporating geometric consistency constraints between surface normals and light direction. Furthermore, we employ a closed-loop feedback mechanism to iteratively refine geometrically consistent shadow maps in self-occluded regions. Experiments demonstrate that our method reduces the Background Error Rate (BER) for attached shadow detection by ≥33% over prior art, while maintaining state-of-the-art performance on cast shadow detection and overall shadow segmentation.

Technology Category

• Computer Vision: Scene Analysis & Understanding
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Attached shadows occur on the surface of the occluder where light cannot reach because of self-occlusion. They are crucial for defining the three-dimensional structure of objects and enhancing scene understanding. Yet existing shadow detection methods mainly target cast shadows, and there are no dedicated datasets or models for detecting attached shadows. To address this gap, we introduce a framework that jointly detects cast and attached shadows by reasoning about their mutual relationship with scene illumination and geometry. Our system consists of a shadow detection module that predicts both shadow types separately, and a light estimation module that infers the light direction from the detected shadows. The estimated light direction, combined with surface normals, allows us to derive a geometry-consistent partial map that identifies regions likely to be self-occluded. This partial map is then fed back to refine shadow predictions, forming a closed-loop reasoning process that iteratively improves both shadow segmentation and light estimation. In order to train our method, we have constructed a dataset of 1,458 images with separate annotations for cast and attached shadows, enabling training and quantitative evaluation of both. Experimental results demonstrate that this iterative geometry-illumination reasoning substantially improves the detection of attached shadows, with at least 33% BER reduction, while maintaining strong full and cast shadow performance.

Problem

• Detects attached shadows using 3D geometry and light direction
• Addresses lack of dedicated datasets for attached shadow detection
• Improves shadow segmentation through iterative illumination-geometry reasoning

Innovation

• Joint detection of cast and attached shadows
• Closed-loop reasoning with light estimation and geometry
• Dataset with separate annotations for both shadow types