Research map / 3D vision / NeRF and radiance fields
MobileNeRF: Exploiting the Polygon Rasterization Pipeline for Efficient Neural Field Rendering on Mobile Architectures
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 polygon rasterization with textured features and a deferred neural shader instead of voxel ray marching | MobileNeRF vs SNeRG · Synthetic 360°, Desktop plugged in | FPS: 207.26 -> 744.91 | Helped |
| Use polygon rasterization with textured features and a deferred neural shader instead of voxel ray marching | MobileNeRF vs SNeRG · Synthetic 360° | GPU memory: 2707.25 -> 538.38 | Helped |
| Use MobileNeRF rather than the strongest listed baseline | NeRF rendering · Synthetic 360° | PSNR: 31.65 -> 30.90 | Hurt |
| Use MobileNeRF rather than the strongest listed baseline | NeRF rendering · Forward-facing | PSNR: 26.92 -> 25.91 | Hurt |
| Optimize mesh vertices rather than fix them on a regular grid (ablation) | MobileNeRF Stage 1 · Synthetic 360° | PSNR: 29.87 -> 32.13 | Helped |
| Include a view-dependent MLP (ablation) | MobileNeRF Stage 1 · Synthetic 360° | PSNR: 29.91 -> 32.13 | Helped |
| Use a smaller grid, P=64 instead of P=128 (ablation) | MobileNeRF Stage 1 · Synthetic 360° | PSNR: 32.13 -> 31.58 | Hurt |
| Remove the acceleration grid (ablation) | MobileNeRF Stage 1 · Synthetic 360° | PSNR: 32.13 -> 31.77 | Hurt |
| Fine-tune feature and shader weights after binarization (ablation) | MobileNeRF Stage 2 · Synthetic 360° | PSNR: 30.80 -> 31.01 | Helped |
| Co-train continuous and binary opacity losses rather than use only binary pseudo-gradients (ablation) | MobileNeRF Stage 2 · Synthetic 360° | PSNR: 29.70 -> 31.01 | Helped |
| Use pseudo-gradients through binary rendering rather than binary opacity loss (ablation) | MobileNeRF Stage 2 · Synthetic 360° | PSNR: 30.89 -> 31.01 | Helped |
| Increase texture size from K=17 to K=33 (ablation) | MobileNeRF Stage 3 · Synthetic 360° | PSNR: 30.90 -> 30.99 | Helped |
Closest papers by meaning
- Real-Time Neural Light Field on Mobile Devices (CVPR 2023 · similarity 0.91)
- VideoRF: Rendering Dynamic Radiance Fields as 2D Feature Video Streams (CVPR 2024 · similarity 0.89)
- A General Implicit Framework for Fast NeRF Composition and Rendering (AAAI 2024 · similarity 0.88)
- Ced-NeRF: A Compact and Efficient Method for Dynamic Neural Radiance Fields (AAAI 2024 · similarity 0.87)
- Rad-NeRF: Ray-decoupled Training of Neural Radiance Field (NeurIPS 2024 · similarity 0.87)
- Learning Robust Generalizable Radiance Field with Visibility and Feature Augmented Point Representation (ICLR 2024 · 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.