Research map / 3D vision / Gaussian splatting
CapFrame: Text-Instructed Viewpoint Grounding in 3D Gaussian Scenes via Geometric Pseudo Labels
Jirong Li, Satoshi Ikehata, Shuhei Kurita, Ikuro Sato
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 camera pose refinement with pseudo-label losses after retrieval (ablation) | CapFrame · Mip-NeRF 360 | SigLIP2: 0.121 -> 0.244 | Improved |
| add camera pose refinement (ablation) | CapFrame · LERF-OVS | SigLIP2: 0.503 -> 0.708 | Improved |
| add camera pose refinement (ablation) | CapFrame · Tanks and Temples | SigLIP2: 0.307 -> 0.339 | Improved |
| CapFrame vs best heuristic baseline SVS | CapFrame · TIVG | GPT alignment score: 4.06 -> 4.75 | Improved |
| remove orientation pseudo label (ablation) | CapFrame · TIVG | rotation deviation / SigLIP2: dR 33.71 deg; dSigLIP2 -0.128 | Hurt |
| remove layout pseudo label (ablation) | CapFrame · TIVG | translation deviation / SigLIP2: dt 5.701; dSigLIP2 -0.236 | Hurt |
| use Top-5 instead of Top-2 retrieved images for initialization (ablation) | CapFrame · TIVG | SigLIP2: dSigLIP2 -0.165 | Hurt |
| paraphrase the input instruction (ablation) | CapFrame · TIVG | SigLIP2: dSigLIP2 +0.032 | No clear effect |
| add moderate noise to pseudo labels (ablation) | CapFrame · TIVG | SigLIP2: dSigLIP2 -0.048 | No clear effect |
| add Question-Evaluation to FG-CLIP retrieval (ablation) | CapFrame · TIVG retrieval | initial view quality | Improved |
| optimize pose at low resolution (ablation) | CapFrame · TIVG runtime | time per query: 102.5 s -> 55.8 s | Improved |
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.
- ChatCam — ChatCam: Empowering Camera Control through Conversational AI (NeurIPS 2024) · better on 5 of 5 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
- PF3plat: Pose-Free Feed-Forward 3D Gaussian Splatting for Novel View Synthesis (ICML 2025 · similarity 0.88)
- SplatFlow: Multi-View Rectified Flow Model for 3D Gaussian Splatting Synthesis (CVPR 2025 · similarity 0.88)
- G^3-LQ: Marrying Hyperbolic Alignment with Explicit Semantic-Geometric Modeling for 3D Visual Grounding (CVPR 2024 · similarity 0.87)
- COLMAP-Free 3D Gaussian Splatting (CVPR 2024 · similarity 0.86)
- Tackling View-Dependent Semantics in 3D Language Gaussian Splatting (ICML 2025 · similarity 0.86)
- SeeGround: See and Ground for Zero-Shot Open-Vocabulary 3D Visual Grounding (CVPR 2025 · similarity 0.86)
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.
