Research map / 3D vision / NeRF and radiance fields
TinyNeRF: Towards 100 x Compression of Voxel Radiance Fields
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 |
|---|---|---|---|
| Compress DVGO voxel grids to 2MB using block-wise DCT, pruning, and quantization | DVGO · Synthetic NeRF | PSNR: 31.95 -> 31.72 | Hurt |
| Compress DVGO voxel grids to 2MB using block-wise DCT, pruning, and quantization | DVGO · Synthetic-NSVF | PSNR: 35.08 -> 34.62 | Hurt |
| Compress DVGO voxel grids to 2MB using block-wise DCT, pruning, and quantization | DVGO · Blended MVS | PSNR: 28.02 -> 28.02 | No clear effect |
| Compress DVGO voxel grids to 2MB using block-wise DCT, pruning, and quantization | DVGO · T&T | PSNR: 28.41 -> 28.19 | Hurt |
| Compress DVGO voxel grids to 2MB using block-wise DCT, pruning, and quantization | DVGO · Synthetic NeRF | model size (MB): 203 -> 2 | Helped |
| Use frequency-domain rather than spatial-domain pruning and quantization (ablation) | TinyNeRF · Synthetic NeRF | PSNR | Helped |
| Apply block-wise DCT instead of whole-grid DCT (ablation) | TinyNeRF · Synthetic NeRF | PSNR | Helped |
| Fine-tune the compressed coefficients after transforming, pruning, and quantizing (ablation) | TinyNeRF · Synthetic NeRF | PSNR: 31.95 -> 31.68 | Helped |
| Vary pruning ratios and quantization bit widths across the reported configurations (ablation) | TinyNeRF · Synthetic NeRF | PSNR: 31.72; 31.59; 31.59; 31.54 | No clear effect |
| Use block-wise DCT, pruning, and quantization at 2MB | DVGO · Synthetic NeRF | training time: 3 minutes -> 6.5 minutes | Hurt |
Closest papers by meaning
- Sparse Voxels Rasterization: Real-time High-fidelity Radiance Field Rendering (CVPR 2025 · similarity 0.91)
- Learning with Unreliability: Fast Few-shot Voxel Radiance Fields with Relative Geometric Consistency (CVPR 2024 · similarity 0.90)
- FrugalNeRF: Fast Convergence for Extreme Few-shot Novel View Synthesis without Learned Priors (CVPR 2025 · similarity 0.90)
- CVT-xRF: Contrastive In-Voxel Transformer for 3D Consistent Radiance Fields from Sparse Inputs (CVPR 2024 · similarity 0.89)
- Spatial Annealing for Efficient Few-shot Neural Rendering (AAAI 2025 · similarity 0.88)
- Neural NeRF Compression (ICML 2024 · similarity 0.88)
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.