Research map / 3D vision / Gaussian splatting
TriNeRFLet: A Wavelet Based Triplane NeRF Representation
Rajaei Khatib, Raja Giryes
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 |
|---|---|---|---|
| learn wavelet coefficients of the triplane instead of raw planes | Triplane · Blender reconstruction | PSNR (avg): 31.26 -> 33.07 (Base Light); 31.64 -> 33.43 (Base, 6h) | Improved |
| add wavelet L1 regularization on LH, HL, HH (ablation) | TriNeRFLet Base Light (no c2f) · Blender reconstruction | PSNR (avg): 31.55 -> 32.93 | Improved |
| add coarse-to-fine training (ablation) | TriNeRFLet Base Light · Blender reconstruction | PSNR (avg): 32.93 -> 33.07 | Improved |
| use multiscale (low-res wavelet levels) LR rendering in SR (ablation) | TriNeRFLetSR · Blender 200->800 | PSNR: 28.49 -> 29.49 | Improved |
| add LPIPS loss between HR render and LR ground truth (ablation) | TriNeRFLetSR · Blender 200->800 | PSNR: 29.27 -> 29.49 | Improved |
| use TriNeRFLet instead of INGP as representation in the SR scheme (ablation) | SR scheme · Blender 200->800 | PSNR: 28.25 -> 29.49 | Improved |
| TriNeRFLetSR (no 3D supervision) vs NVSR (3D supervised) | Blender 200->800 SR · Blender 200->800 | PSNR: 29.53 -> 29.49 | No clear effect |
| TriNeRFLetSR vs NeRF-SR (SS) on LLFF | LLFF SR · LLFF 4x | PSNR: 25.13 -> 25.00 | Hurt |
| use Haar wavelet instead of Bior6.8 (ablation) | TriNeRFLet · Blender 4 scenes | training time and quality: up to 30% faster training | Mixed |
| wavelet reconstruction per training step | TriNeRFLet · Blender | train time: INGP 5 mins vs TriNeRFLet Small 15 mins | Hurt |
Closest papers by meaning
- WaveNeRF: Wavelet-based Generalizable Neural Radiance Fields (ICCV 2023 · similarity 0.88)
- Nerflets: Local Radiance Fields for Efficient Structure-Aware 3D Scene Representation From 2D Supervision (CVPR 2023 · similarity 0.88)
- Tetra-NeRF: Representing Neural Radiance Fields Using Tetrahedra (ICCV 2023 · similarity 0.88)
- Improving the Convergence of Dynamic NeRFs via Optimal Transport (ICLR 2024 · similarity 0.88)
- CodecNeRF: Toward Fast Encoding and Decoding, Compact, and High-quality Novel-view Synthesis (AAAI 2025 · similarity 0.87)
- ProvNeRF: Modeling per Point Provenance in NeRFs as a Stochastic Field (NeurIPS 2024 · 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.
