Research map / 3D vision / Face and shape
Streamable Neural Fields
Junwoo Cho, Seungtae Nam, Daniel Rho, Jong Hwan Ko, Eunbyung Park
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 |
|---|---|---|---|
| use progressive growing with weight reuse instead of slimmable training (ablation) | streamable neural field · Kodak spectral growing, sub-network 4 | PSNR: 29.19 -> 30.91 | Improved |
| use progressive growing instead of slimmable training (ablation) | streamable neural field · UVG temporal growing, sub-network 3 | PSNR: 34.54 -> 38.22 | Improved |
| use progressive growing instead of slimmable training (ablation) | streamable neural field · Kodak spatial growing, sub-network 1 | PSNR: 24.49 -> 30.04 | Improved |
| train streamable progressive model instead of individual models | streamable neural field · Kodak spectral growing, sub-network 3 | PSNR: 29.78 -> 29.29 | Hurt |
| train streamable progressive model instead of individual models | streamable neural field · Kodak spectral growing, sub-network 3 | LPIPS: 0.246 -> 0.190 | Improved |
| train streamable progressive model instead of individual models | streamable neural field · Kodak spatial growing, sub-network 2 | PSNR: 28.82 -> 29.62 | Improved |
| train streamable progressive model instead of individual models | streamable neural field · UVG temporal growing, sub-network 3 | PSNR: 37.60 -> 38.22 | Improved |
| train streamable progressive model instead of individual models | streamable neural field · 3D spectral growing armadillo, sub-network 2 | Chamfer distance: 0.137 -> 0.032 | Improved |
| share parameters across sub-networks instead of separate models | streamable neural field · Kodak image 23 parameter count | parameters: 590K -> 195K | Improved |
| use proposed weight initialization instead of SIREN initialization for newly added weights (ablation) | streamable neural field · Kodak (averaged PSNR curve) | PSNR | Improved |
| use proposed initialization instead of SIREN initialization (ablation) | streamable neural field · activation and gradient statistics | activation/gradient distribution | No clear effect |
| share parameters via progressive growing | streamable neural field · Kodak image 19 loss surface | training stability | Improved |
Closest papers by meaning
- How to Train Neural Field Representations: A Comprehensive Study and Benchmark (CVPR 2024 · similarity 0.86)
- ResFields: Residual Neural Fields for Spatiotemporal Signals (ICLR 2024 · similarity 0.86)
- Neural Fields as Distributions: Signal Processing Beyond Euclidean Space (CVPR 2024 · similarity 0.85)
- BANF: Band-Limited Neural Fields for Levels of Detail Reconstruction (CVPR 2024 · similarity 0.84)
- From Activation to Initialization: Scaling Insights for Optimizing Neural Fields (CVPR 2024 · similarity 0.83)
- Coordinate-Aware Modulation for Neural Fields (ICLR 2024 · similarity 0.83)
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.
