PRISM: Learning Realistic Depth via Physics-Grounded Noise Disentanglement with Semantic-Geometric Collaboration
PRISM: Learning Realistic Depth via Physics-Grounded Noise Disentanglement with Semantic-Geometric Collaboration
PRISM is a tripartite framework that disentangles depth noise into sensing invalidation and measurement inaccuracy, enabling physics-based reasoning. It achieves state-of-the-art fidelity in noisy depth synthesis and a 92.1% success rate in real-world robotic tasks through hierarchical positive-prioritized supervision.
This paper introduces PRISM, a novel framework for depth sensing that disentangles noise into physical modalities, achieving state-of-the-art performance in noisy depth synthesis and a 92.1% success rate in real-world robotic tasks.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
noise modalities
2
View original evidence
k, we introduce a physics-grounded paradigm that disentangles monolithic noise into two complementary modalities: \textbf{sensing invalidation} and \textbf{measurement inaccuracy}, enabling a tailored treatment of noise sources based on their physical origins. Building on this insight, we propose \textbf{PRISM}, a tripartite framework that distills 3D Visual Foundation Model features as rich spatial-semantic priors f
Open source S3
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
urthermore, downstream robotic experiments show that PRISM facilitates a \textbf{92.1\%} average success rate in the real world, marking a significant improvement over monolithic baselines. Our project page is available at: \resizebox{\linewidth}{!}{\href{https://shawnleung-sii.github.io/prism-icml2026/}{https://shawnleung-sii.github.io/prism-icml2026/}.}
Open source S5
or physics-based reasoning. To address the inherent sparsity and class imbalance of invalidation regions, we develop Hierarchical Positive-Prioritized Supervision, integrating multi-scale positive-weighted objectives with a positive-preserving dynamic hard mining strategy to enforce precise artifact delineation. Extensive benchmarks demonstrate that PRISM achieves state-of-the-art fidelity in noisy depth synthesis. F
Open source S4
k, we introduce a physics-grounded paradigm that disentangles monolithic noise into two complementary modalities: \textbf{sensing invalidation} and \textbf{measurement inaccuracy}, enabling a tailored treatment of noise sources based on their physical origins. Building on this insight, we propose \textbf{PRISM}, a tripartite framework that distills 3D Visual Foundation Model features as rich spatial-semantic priors f
Open source S3
Source：arXiv Robotics — research abstracts · arxiv.org