PlenoCI: Plenoptic CharacterIstics for View Dependence Aware Change Classification

2026-09-23

AI Summary

This study addresses the challenge that under-constrained radiance field reconstruction causes independently optimized models to produce spurious differences, hindering accurate change detection. To overcome this, we propose PlenoCI features, which for the first time derive closed-form plenoptic derivatives from 3D Gaussian Splatting (3DGS). By leveraging view-dependent behavior to capture changes while suppressing diffuse texture interference, this approach structurally eliminates under-constrained noise. Experimental results on the CL-Splats benchmark demonstrate that PlenoCI improves mean Intersection over Union (mIoU) by 25.7%, reduces false positive rates in unchanged regions by two orders of magnitude, and achieves a balanced accuracy of 0.735 for geometry and appearance classification, thereby enabling robust instance-aware change detection.

Abstract

Radiance field representations such as 3D Gaussian Splatting (3DGS) natively encode complex visual phenomena such as occlusions and view dependence, but they are inherently underconstrained. Independently optimized reconstructions converge to different primitive configurations, even in unchanged regions. We introduce Plenoptic CharacterIstics (PlenoCI), a novel feature built from the plenoptic field these representations approximate. PlenoCI directly captures rich visual behaviors while ignoring Lambertian textures. By deriving closed-form analytic plenoptic derivatives from a 3DGS representation, we efficiently detect these 5D structures. Our approach is robust to underconstrained representations by construction, reporting two orders of magnitude fewer false positives between independent reconstructions of unchanged scenes than concurrent work. We demonstrate PlenoCI's utility on change classification. First, we detect changes with an instance-aware 3DGS pipeline, achieving state-of-the-art results on CL-Splats with a 25.7% mIoU gain over the strongest competitor, while remaining competitive on the more challenging PASLCD benchmark. Leveraging PlenoCI, we classify changes as geometric or appearance-based with a balanced accuracy of 0.735, comparable to the best performing baseline. We believe plenoptic derivatives and PlenoCI open new directions for view dependence aware understanding in visually complex environments. Code and data are available at https://js0n-lai.github.io/plenoci.

Problem

• Change Classification
• View Dependence
• 3D Gaussian Splatting
• Radiance Fields
• Plenoptic Characteristics

Innovation

• Plenoptic Characteristics
• 3D Gaussian Splatting
• Change Classification
• Analytic Plenoptic Derivatives
• View Dependence