Research map / 3D vision / Gaussian splatting
IRIS: Intersection-aware Ray-based Implicit Editable Scenes
Grzegorz Wilczyński, Mikołaj Zieliński, Krzysztof Byrski, Joanna Waczynska, Dominik Belter, Przemysław Spurek
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 |
|---|---|---|---|
| increase max intersections per ray from 16 to 128 (ablation) | IRIS · NeRF Synthetic ablation (avg) | PSNR: 29.87 -> 32.65 | Improved |
| increase max intersections per ray from 64 to 128 (ablation) | IRIS · NeRF Synthetic ablation (avg) | PSNR: 32.47 -> 32.65 | Improved |
| add multi-resolution hash grid feature initialization (ablation) | IRIS · NeRF Synthetic ablation (avg) | PSNR: 32.54 -> 32.65 | Improved |
| use analytical ray-intersection sampling instead of EKS grid sampling | IRIS · NeRF Synthetic (average) | FPS: 0.183 -> 6.038 | Improved |
| use IRIS instead of EKS for training | IRIS · NeRF Synthetic (average) | training time: 26h 2min -> 51min 18s | Improved |
| use IRIS instead of EKS | IRIS · NeRF Synthetic (average) | PSNR: 33.12 -> 32.65 | Hurt |
| use IRIS instead of EKS on outdoor scenes | IRIS · Mip-NeRF 360 outdoor | PSNR: 20.70 -> 24.11 | Improved |
| use IRIS instead of Radiance Meshes on outdoor scenes | IRIS · Mip-NeRF 360 outdoor | PSNR: 24.73 -> 24.11 | Hurt |
| use IRIS instead of Radiance Meshes on indoor scenes | IRIS · Mip-NeRF 360 indoor | PSNR: 29.29 -> 30.85 | Improved |
| use IRIS instead of EKS on deformation benchmark | IRIS · NeuralEditor deformation benchmark (avg) | PSNR: 25.92 -> 25.70 | No clear effect |
| use IRIS instead of NeuralEditor on deformation benchmark | IRIS · NeuralEditor deformation benchmark (avg) | PSNR: 25.12 -> 25.70 | Improved |
| use ray intersection selector instead of proposal or volumetric sampler at large batch | IRIS · Sampler throughput | scalability | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- Mip-NeRF 360 — Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fields (CVPR 2022) · better on 0 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
- Learning Unified Decompositional and Compositional NeRF for Editable Novel View Synthesis (ICCV 2023 · similarity 0.87)
- Learning Object-Compositional Neural Radiance Field for Editable Scene Rendering (ICCV 2021 · similarity 0.87)
- Feature 3DGS: Supercharging 3D Gaussian Splatting to Enable Distilled Feature Fields (CVPR 2024 · similarity 0.85)
- Structure Consistent Gaussian Splatting with Matching Prior for Few-shot Novel View Synthesis (NeurIPS 2024 · similarity 0.85)
- SIGNeRF: Scene Integrated Generation for Neural Radiance Fields (CVPR 2024 · similarity 0.84)
- LAENeRF: Local Appearance Editing for Neural Radiance Fields (CVPR 2024 · similarity 0.84)
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.
