Research map / 3D vision / NeRF and radiance fields
Free View Synthesis
Gernot Riegler, Vladlen Koltun
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 source image encoding before mapping (ablation) | recurrent blending network · Tanks and Temples Truck (leave-one-out, quarter res) | LPIPS: 0.093 -> 0.082 | Improved |
| add source image encoding before mapping (ablation) | recurrent blending network · Tanks and Temples Train (leave-one-out, quarter res) | LPIPS: 0.174 -> 0.147 | Improved |
| add source image encoding before mapping (ablation) | recurrent blending network · Tanks and Temples Truck (leave-one-out, quarter res) | PSNR: 22.13 -> 22.03 | Hurt |
| add GRU recurrent units to blending decoder (ablation) | blending network · Tanks and Temples Train (leave-one-out, quarter res) | LPIPS: 0.159 -> 0.147 | Improved |
| add GRU recurrent units to blending decoder (ablation) | blending network · Tanks and Temples Truck (leave-one-out, quarter res) | LPIPS: 0.094 -> 0.082 | Improved |
| append mapping masks to blending input (ablation) | blending network · Tanks and Temples Train (leave-one-out, quarter res) | LPIPS: 0.152 -> 0.147 | Improved |
| append mapping masks to blending input (ablation) | blending network · Tanks and Temples Truck (leave-one-out, quarter res) | PSNR: 21.58 -> 22.03 | Improved |
| warp invalid depth values with infinite depth (ablation) | blending network · Tanks and Temples Train (leave-one-out, quarter res) | LPIPS: 0.169 -> 0.147 | Improved |
| add soft-argmax over per-image outputs (ablation) | blending network · Tanks and Temples Train (leave-one-out, quarter res) | LPIPS: 0.159 -> 0.147 | Improved |
| add VGG encoding to fixed-input network (ablation) | Fixed Identity · Tanks and Temples Truck (leave-one-out, quarter res) | LPIPS: 0.116 -> 0.096 | Improved |
| replace global-average concatenation with recurrent GRU blending (ablation) | Cat Global Avg. · Tanks and Temples Train (leave-one-out, quarter res) | LPIPS: 0.175 -> 0.147 | Improved |
| increase number of source images beyond 7 (ablation) | recurrent blending network · Tanks and Temples | image fidelity | No clear effect |
Closest papers by meaning
- LVSM: A Large View Synthesis Model with Minimal 3D Inductive Bias (ICLR 2025 · similarity 0.90)
- Novel View Synthesis with Pixel-Space Diffusion Models (CVPR 2025 · similarity 0.90)
- Free3D: Consistent Novel View Synthesis without 3D Representation (CVPR 2024 · similarity 0.88)
- Less Decoder is More Encoder: Geometric Representation Learning from Novel View Synthesis (NeurIPS 2026 · similarity 0.88)
- MVSplat360: Feed-Forward 360 Scene Synthesis from Sparse Views (NeurIPS 2024 · similarity 0.88)
- GenRec: Knowing Where to Reconstruct and Where to Generate (NeurIPS 2026 · similarity 0.87)
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.
