Benchmark reveals key factors for effective 3D scene editing
What Makes a 3D Scene Editable? A Factorized Benchmark of Fidelity, Locality, Consistency, and Preservation
Computer Vision and Pattern Recognition
Summary
Editing 3D scenes in computer graphics often looks good when focusing only on matching what you want changed, but this can accidentally alter other parts or cause glitches when viewed from different angles. The authors created a new test called EditBench3D that checks how well edits follow instructions, stay local, look consistent from all views, and preserve unchanged parts. They tested eight different editing methods on various changes and found that doing well on one aspect doesn’t guarantee quality in others. Their results suggest it's better to measure editing by several qualities rather than a single score.
What this means in practice
- •For 3d artists and modelers: Evaluate and choose 3D editing tools that balance accuracy, detail isolation, and view consistency for complex scene modifications.
- •For game developers: Improve in-game content editing by understanding trade-offs between edit fidelity and consistency across different viewpoints.
Authors
Sariah Patro, Arjun Mehra, Nikhil Bhatia
Abstract
Neural 3D scene editing is often evaluated by semantic alignment alone, although a convincing result may alter unrelated content or become inconsistent across views. We introduce EditBench3D, a representation-agnostic benchmark that treats editing as controlled information replacement. It evaluates four complementary properties: instruction fidelity, spatial locality, cross-view consistency, and preservation of non-target content. The protocol combines visibility-aware 3D target supports, paired descriptions, held-out cameras, and five edit families covering appearance, material, geometry, and object-level changes. We evaluate eight representative NeRF, 3D Gaussian Splatting, hybrid, and proxy-based editors on 240 scene-edit pairs. The study shows that semantic fidelity is only weakly associated with the other editing properties, and that no single method is optimal across all dimensions. Explicit Gaussian editors offer a strong overall balance, whereas direct proxy manipulation provides the most conservative edits at the cost of open-ended fidelity. These findings support reporting editability as a multi-objective profile rather than a single semantic score.