NTIRE 2025 Image Shadow Removal Challenge Report

2025-06-18

AI Summary

This work addresses the challenging problem of shadow removal in real-world scenes, where self-shadows and cast shadows are tightly coupled, severely degrading restoration fidelity. To systematically evaluate solutions, we propose a dual-track benchmark framework—built upon the WSRD+ dataset—that jointly quantifies pixel-level reconstruction accuracy and human visual perception, establishing the first unified evaluation standard for both metrics. Our method employs an end-to-end deep network integrating multi-scale feature extraction, shadow-mask-guided attention, and a perception-driven loss function. As the core organizing effort of the NTIRE 2025 Image Shadow Removal Challenge, it attracted 306 registered teams, with 17 advancing to the final round. Comprehensive evaluation reveals persistent limitations in handling complex material reflectance and heavily overlapping shadow regions, highlighting critical bottlenecks and underscoring the necessity of shifting from reconstruction-oriented to perception-oriented paradigms in shadow removal research.

Technology Category

• Computer Vision: Motion & Tracking
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Security and Privacy: Large-scale security measurements

Abstract

This work examines the findings of the NTIRE 2025 Shadow Removal Challenge. A total of 306 participants have registered, with 17 teams successfully submitting their solutions during the final evaluation phase. Following the last two editions, this challenge had two evaluation tracks: one focusing on reconstruction fidelity and the other on visual perception through a user study. Both tracks were evaluated with images from the WSRD+ dataset, simulating interactions between self- and cast-shadows with a large number of diverse objects, textures, and materials.

Problem

• Removing shadows from images accurately
• Evaluating fidelity and visual perception
• Handling diverse objects and textures

Innovation

• Large-scale WSRD+ dataset for evaluation
• Dual-track fidelity and perception assessment
• 306 participants with 17 final solutions