Research map / 3D vision / Gaussian splatting
Geometry-Aware Cross-Modal Graph Alignment for Referring Segmentation in 3D Gaussian Splatting
Yuwen Tao, Kanglei Zhou, Chang Li, Liyuan Wang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add semantic graph (relation-aware text graph) (ablation) | GeoCGA baseline · Ref-LERF Ramen / Kitchen | mIoU: 28.3 -> 29.5 (Ramen); 20.1 -> 23.8 (Kitchen) | Improved |
| add 3D geometry graph (ablation) | GeoCGA baseline · Ref-LERF Ramen / Kitchen | mIoU: 28.3 -> 30.4 (Ramen); 20.1 -> 26.5 (Kitchen) | Improved |
| combine semantic and geometry graphs (ablation) | GeoCGA · Ref-LERF Ramen / Kitchen | mIoU: 28.3 -> 32.1 (Ramen); 20.1 -> 30.3 (Kitchen) | Improved |
| use edge-aware message passing instead of plain semantic GNN (ablation) | GeoCGA · Ref-LERF Ramen / Kitchen | mIoU: 31.5 -> 32.1 (Ramen); 29.1 -> 30.3 (Kitchen) | Improved |
| use cross-view geometric consistency loss instead of pseudo-mask loss (ablation) | GeoCGA · Ref-LERF Ramen / Kitchen | mIoU: 30.9 -> 32.1 (Ramen); 29.4 -> 30.3 (Kitchen) | Improved |
| add explicit relation-level matching (Eq. 10) (ablation) | GeoCGA · Ref-LERF Ramen / Kitchen | mIoU: 31.1 -> 32.1 (Ramen); 29.5 -> 30.3 (Kitchen) | Improved |
| replace ReferSplat with GeoCGA | ReferSplat · Ref-LERF (avg) | mIoU: 25.0 -> 30.2 | Improved |
| replace ReferSplat with GeoCGA | ReferSplat · LERF-OVS (avg) | mIoU: 52.6 -> 55.6 | Improved |
| replace ReferSplat with GeoCGA | ReferSplat · 3D-OVS (avg) | mIoU: 92.9 -> 93.7 | Improved |
| GeoCGA vs LangSplat on Figurines scene | GeoCGA · LERF-OVS Figurines | mIoU: 65.1 -> 63.1 (vs LangSplat) | Hurt |
| use object-level graphs instead of dense Gaussian features | ReferSplat · model size | trainable parameters (M): 304.18 -> 128.48 | Improved |
| use object-level graphs and lightweight GCN alignment | ReferSplat · inference cost | FLOPs (G): 41.82 -> 25.28 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- LangSplat — LangSplat: 3D Language Gaussian Splatting (CVPR 2024) · better on 2 of 3 shared results
- ReferSplat — ReferSplat: Referring Segmentation in 3D Gaussian Splatting (ICML 2025) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- ReferSplat: Referring Segmentation in 3D Gaussian Splatting (ICML 2025 · similarity 0.91)
- X-RefSeg3D: Enhancing Referring 3D Instance Segmentation via Structured Cross-Modal Graph Neural Networks (AAAI 2024 · similarity 0.89)
- 3D Vision-Language Gaussian Splatting (ICLR 2025 · similarity 0.88)
- Chain of Semantics Programming in 3D Gaussian Splatting Representation for 3D Vision Grounding (CVPR 2025 · similarity 0.88)
- Bootstraping Clustering of Gaussians for View-consistent 3D Scene Understanding (AAAI 2025 · similarity 0.88)
- PanoGS: Gaussian-based Panoptic Segmentation for 3D Open Vocabulary Scene Understanding (CVPR 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
