NTIRE 2025 Challenge on HR Depth from Images of Specular and Transparent Surfaces

2025-06-06

AI Summary

This work addresses two key challenges in high-resolution monocular/stereo depth estimation on non-Lambertian surfaces—such as specular and transparent objects: resolution limitations and inaccurate material modeling. We propose the first high-resolution depth estimation benchmark and methodological framework explicitly designed for non-Lambertian materials. Our contribution comprises (1) a novel dual-track evaluation benchmark—supporting both single-image and stereo-depth settings—to fill the gap in standardized assessment for this domain; and (2) an end-to-end deep learning model integrating multi-scale feature extraction, material-aware attention, geometric consistency constraints, and differentiable rendering priors. Among 177 registered teams, eight submitted complete solutions. The top-performing method achieves a 3.2 dB PSNR improvement on high-resolution depth maps and a 37% reduction in boundary error for specular and transparent objects—significantly enhancing robustness and accuracy of depth perception in complex optical scenes.

Technology Category

• Computer Vision: 3D Computer Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Knowledge Representation and Reasoning: Nonmonotonic Reasoning

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Security and Privacy: Data transparency and provenance

Abstract

This paper reports on the NTIRE 2025 challenge on HR Depth From images of Specular and Transparent surfaces, held in conjunction with the New Trends in Image Restoration and Enhancement (NTIRE) workshop at CVPR 2025. This challenge aims to advance the research on depth estimation, specifically to address two of the main open issues in the field: high-resolution and non-Lambertian surfaces. The challenge proposes two tracks on stereo and single-image depth estimation, attracting about 177 registered participants. In the final testing stage, 4 and 4 participating teams submitted their models and fact sheets for the two tracks.

Problem

• Advance depth estimation for high-resolution images
• Address depth estimation for non-Lambertian surfaces
• Propose stereo and single-image depth estimation tracks

Innovation

• High-resolution depth estimation from images
• Focus on specular and transparent surfaces
• Dual tracks for stereo and single-image methods