Paper · ≈ 1 min read
MERF: Memory-Efficient Radiance Fields for Real-time View Synthesis in Unbounded Scenes
Outgoing relations
- evaluated_on → Mip-NeRF 360 (Benchmark)
- part_of → 3D/4D Vision and Reconstruction (ResearchField)
- part_of → Research Literature (ResearchField)
- uses → Volumetric Rendering (MethodologicalConcept)
- uses_metric → LPIPS (Metric)
- uses_metric → PSNR (Metric)
- uses_metric → SSIM (Metric)
Connected node types
- Metric: 3
- ResearchField: 2
- Benchmark: 1
- MethodologicalConcept: 1
Mentions in the corpus
- part_of Evidence: MERF reduces the memory consumption of prior sparse volumetric radiance…
- mentioned_in Claim: MERF reduces the memory consumption of prior sparse volumetric radiance fields using a combinat…
- part_of Evidence: We present a Memory-Efficient Radiance Field (MERF) representation that…
- part_of Evidence: To support large-scale unbounded scenes, we introduce a novel contracti…
- part_of Evidence: Neural radiance fields enable state-of-the-art photorealistic view synt…
- part_of Evidence: We design a lossless procedure for baking the parameterization used dur…
- part_of Evidence: A coarse sparse 3D feature grid stores volumetric content near surfaces…
- mentioned_in Claim: A coarse sparse 3D feature grid stores volumetric content near surfaces, with empty space prune…
- summarizes Project Pulse
- summarizes Field Overview — 3D/4D Vision and Reconstruction
- summarizes Field Overview — Research Literature
Cross-references in raw data
Source provenance
examples/demo-corpus/data/research/papers/arxiv-2302-12249/abstract.md