![[2602.11638] Variation-aware Flexible 3D Gaussian Editing](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.11638] Variation-aware Flexible 3D Gaussian Editing
Summary
The paper presents VF-Editor, a novel approach for flexible 3D Gaussian editing that addresses limitations of indirect editing methods by enabling direct attribute variation predictions.
Why It Matters
This research is significant as it improves the efficiency and flexibility of 3D editing processes, which are crucial for applications in graphics and AI. By overcoming the limitations of existing methods, VF-Editor could enhance workflows in various fields, including game design and virtual reality.
Key Takeaways
- VF-Editor allows for native editing of Gaussian primitives in 3D.
- It predicts attribute variations using a novel variation predictor based on 2D editing knowledge.
- The approach addresses cross-view inconsistencies found in indirect editing methods.
- Extensive experiments validate the effectiveness and flexibility of VF-Editor.
- The unified design facilitates knowledge transfer from 2D to 3D editing.
Computer Science > Graphics arXiv:2602.11638 (cs) [Submitted on 12 Feb 2026 (v1), last revised 13 Feb 2026 (this version, v2)] Title:Variation-aware Flexible 3D Gaussian Editing Authors:Hao Qin, Yukai Sun, Meng Wang, Ming Kong, Mengxu Lu, Qiang Zhu View a PDF of the paper titled Variation-aware Flexible 3D Gaussian Editing, by Hao Qin and 5 other authors View PDF HTML (experimental) Abstract:Indirect editing methods for 3D Gaussian Splatting (3DGS) have recently witnessed significant advancements. These approaches operate by first applying edits in the rendered 2D space and subsequently projecting the modifications back into 3D. However, this paradigm inevitably introduces cross-view inconsistencies and constrains both the flexibility and efficiency of the editing process. To address these challenges, we present VF-Editor, which enables native editing of Gaussian primitives by predicting attribute variations in a feedforward manner. To accurately and efficiently estimate these variations, we design a novel variation predictor distilled from 2D editing knowledge. The predictor encodes the input to generate a variation field and employs two learnable, parallel decoding functions to iteratively infer attribute changes for each 3D Gaussian. Thanks to its unified design, VF-Editor can seamlessly distill editing knowledge from diverse 2D editors and strategies into a single predictor, allowing for flexible and effective knowledge transfer into the 3D domain. Extensive experiment...