Research map / 3D vision / Gaussian splatting
GAGS: Granularity-Aware Feature Distillation for Language Gaussian Splatting
Yuning Peng, Haiping Wang, Yuan Liu, Chenglu Wen, Zhen Dong, Bisheng Yang
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 granularity-aware distillation (GaD) (ablation) | GAGS baseline · Mip-NeRF360 (4 annotated scenes) | mAcc: 76.84 -> 86.27 | Improved |
| add granularity-aware distillation (GaD) (ablation) | GAGS baseline · Mip-NeRF360 (4 annotated scenes) | mIoU: 53.72 -> 60.99 | Improved |
| add granularity-aware segmentation (GaS) on top of GaD (ablation) | GAGS · Mip-NeRF360 (4 annotated scenes) | mIoU: 60.99 -> 65.09 | Improved |
| add GaS on top of GaD (ablation) | GAGS · Mip-NeRF360 (4 annotated scenes) | mAcc: 86.27 -> 88.67 | Improved |
| distill only subpart granularity instead of adaptive selection (ablation) | GAGS · Mip-NeRF360 (4 annotated scenes) | mIoU: 65.09 -> 47.78 | Hurt |
| distill only part granularity instead of adaptive selection (ablation) | GAGS · Mip-NeRF360 (4 annotated scenes) | mIoU: 65.09 -> 61.68 | Hurt |
| distill only whole granularity instead of adaptive selection (ablation) | GAGS · Mip-NeRF360 (4 annotated scenes) | mIoU: 65.09 -> 58.63 | Hurt |
| average CLIP features across the three granularities (ablation) | GAGS with GaS · Mip-NeRF360 (4 annotated scenes) | mIoU: 65.09 -> 59.45 | Hurt |
| use one unified feature field with adaptive granularity instead of three fields | LangSplat · ramen (LERF) | inference time (s): 49 -> 24 | Improved |
| use unified field instead of LangSplat's multiple fields | LangSplat · LERF (4 scenes) | overall mAcc: 70.49 -> 79.14 | Improved |
| use unified field instead of LangSplat's multiple fields | LangSplat · Mip-NeRF360 (4 annotated scenes) | overall mAcc: 75.48 -> 88.67 | Improved |
| self-supervised granularity selection | GAGS · complex scenes (failure cases) | component-level recognition | Hurt |
Later papers that compared with this method
4 later accepted papers report GAGS as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- REALM: An MLLM-Agent Framework for Open World 3D Reasoning Segmentation and Editing on Gaussian Splatting (CVPR 2026 · better on 5 of 5 shared results)
- DySurface: Consistent 4D Surface Reconstruction via Bridging Explicit Gaussians and Implicit Functions (NeurIPS 2026 · better on 3 of 4 shared results)
- LEGO: Leveled Language Gaussian Splatting (ECCV 2026 · better on 1 of 1 shared result)
- LESV:Language Embedded Sparse Voxel Fusion for Open-Vocabulary 3D Scene Understanding (ECCV 2026 · better on 1 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- LangSplat — LangSplat: 3D Language Gaussian Splatting (CVPR 2024) · better on 8 of 8 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
- Segment then Splat: Unified 3D Open-Vocabulary Segmentation via Gaussian Splatting (NeurIPS 2025 · similarity 0.90)
- Tackling View-Dependent Semantics in 3D Language Gaussian Splatting (ICML 2025 · similarity 0.90)
- SceneSplat++: A Large Dataset and Comprehensive Benchmark for Language Gaussian Splatting (NeurIPS 2025 · similarity 0.89)
- Dr. Splat: Directly Referring 3D Gaussian Splatting via Direct Language Embedding Registration (CVPR 2025 · similarity 0.89)
- FHGS: Feature-Homogenized Gaussian Splatting (NeurIPS 2025 · similarity 0.89)
- OpenGaussian: Towards Point-Level 3D Gaussian-based Open Vocabulary Understanding (NeurIPS 2024 · similarity 0.89)
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.
