Research map / Image generation / Super-resolution
SIGNET: Efficient Neural Representation for Light Fields
Brandon Yushan Feng, Amitabh Varshney
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 Gegenbauer input transformation on top of sine-activation MLP (SIREN vs SIGNET) (ablation) | SIREN · Stanford static light fields | reconstruction quality | Improved |
| add Gegenbauer input transformation to a ReLU MLP (CoordinateReLU vs GegenbauerReLU) (ablation) | CoordinateReLU · Stanford static light fields | reconstruction quality | Improved |
| replace Fourier input transformations with Gegenbauer (GegenbauerReLU vs FourierReLU, DiscreteFourierReLU, DiscreteFourierSine) (ablation) | MLP light field representation · Stanford static light fields (3 scenes average) | reconstruction quality | Improved |
| use Gegenbauer transformation instead of Fourier-style ones (ablation) | SIGNET · Stanford static light fields | training convergence | Improved |
| increase number of Gegenbauer basis functions C (beta 0.25, 0.5, 2, 4) (ablation) | SIGNET · Lego | PSNR: gain diminishes after C > 1024 | No clear effect |
| increase number of layers L (ablation) | SIGNET · Lego | PSNR: saturates at 10 layers | Improved |
| increase hidden width M (ablation) | SIGNET · Lego | PSNR | No clear effect |
| increase Gegenbauer alpha above default 0.5 (ablation) | SIGNET · Lego | PSNR: alpha larger than 0.5 declines | Hurt |
| add quantization and weight pruning to SIGNET weights | SIGNET · light field compression | bitrate and PSNR: 15-fold bitrate reduction, PSNR decreased by less than 0.1 | Improved |
| use SIGNET instead of AMDE dictionary compression | AMDE · Lego | storage size: 29.3 -> 9.0 | Improved |
Closest papers by meaning
- A Generative Model for Generic Light Field Reconstruction (TPAMI 2020 · similarity 0.88)
- Light Field Neural Rendering (CVPR 2022 · similarity 0.87)
- Learning Neural Light Fields With Ray-Space Embedding (CVPR 2022 · similarity 0.87)
- Light Field Super-Resolution With Zero-Shot Learning (CVPR 2021 · similarity 0.87)
- NeLF-Pro: Neural Light Field Probes for Multi-Scale Novel View Synthesis (CVPR 2024 · similarity 0.87)
- A Pre-convolved Representation for Plug-and-Play Neural Illumination Fields (AAAI 2024 · similarity 0.85)
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.
