Research map / 3D vision / NeRF and radiance fields
Real-Time Neural Light Field on Mobile Devices
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 MobileR2L instead of MobileNeRF | neural rendering on mobile devices · Synthetic 360° | PSNR: 30.90 -> 31.34 | Helped |
| Use MobileR2L instead of MobileNeRF | neural rendering on mobile devices · Forward-facing | PSNR: 25.91 -> 26.15 | Helped |
| Replace MLP layers with 1×1 convolution layers (ablation) | MobileR2L backbone · Chair, synthetic 360° | latency (ms, iPhone 13): 19.57 -> 14.30 | Helped |
| Replace MLP layers with 1×1 convolution layers (ablation) | MobileR2L backbone · Chair, synthetic 360° | PSNR: 19.13 -> 19.16 | No clear effect |
| Use GeLU instead of ReLU (ablation) | MobileR2L · Chair, synthetic 360° | PSNR: 26.82 -> 26.99 | Helped |
| Add batch normalization (ablation) | MobileR2L with GeLU · Chair, synthetic 360° | PSNR: 26.99 -> 27.18 | Helped |
| Reduce input resolution from 100×100 to 50×50 (ablation) | MobileR2L with NeRF teacher · Chair, synthetic 360° | PSNR: 30.65 -> 30.40 | Hurt |
| Increase input ray resolution from 100×100 to 200×200 (ablation) | MobileR2L with NeRF teacher · Chair, synthetic 360° | latency (ms, iPhone 13): 26.21 -> 73.76 | Hurt |
| Use super-resolution modules to render high-resolution images from lower-resolution rays (ablation) | MobileR2L · Chair, synthetic 360° | latency / successful compilation | Helped |
| Use MipNeRF instead of NeRF as teacher (ablation) | MobileR2L · Chair, synthetic 360° | PSNR: 30.65 -> 30.83 | Helped |
| Increase ray samples and positional encoding from K=8, L=6 to K=16, L=10 (ablation) | MobileR2L with MipNeRF teacher · Chair, synthetic 360° | PSNR: 30.83 -> 30.90 | Helped |
| Increase backbone depth to D100 (ablation) | MobileR2L · Chair, synthetic 360° | latency (ms, iPhone 13): 26.21 -> 36.55 | Hurt |
Closest papers by meaning
- VideoRF: Rendering Dynamic Radiance Fields as 2D Feature Video Streams (CVPR 2024 · similarity 0.88)
- A Pre-convolved Representation for Plug-and-Play Neural Illumination Fields (AAAI 2024 · similarity 0.87)
- NeLF-Pro: Neural Light Field Probes for Multi-Scale Novel View Synthesis (CVPR 2024 · similarity 0.87)
- Ced-NeRF: A Compact and Efficient Method for Dynamic Neural Radiance Fields (AAAI 2024 · similarity 0.86)
- A General Implicit Framework for Fast NeRF Composition and Rendering (AAAI 2024 · similarity 0.86)
- CodecNeRF: Toward Fast Encoding and Decoding, Compact, and High-quality Novel-view Synthesis (AAAI 2025 · similarity 0.86)
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.