TetraCoder: adaptive tetrahedral encoding beats Instant-NGP PSNR with fewer parameters
ssh4net · x · 2026-11-16
A new paper introduces TetraCoder, a single-resolution, hash-free adaptive tetrahedral encoding that links an explicit tetrahedral spatial encoder with an underlying implicit signed distance field (SDF).
- The hybrid implicit-explicit space is continuous piecewise affine (CPWA) by design, giving an exact correspondence between the implicit SDF and an explicit analytic mesh, enabling fast analytic mesh extraction
- Unlike Instant-NGP-style multiresolution hash encodings, it builds on an adaptive tetrahedral volumetric mesh with a piecewise-affine decoder
- Includes tetrahedral ray tracing: samples locate their containing cell via an adjacency walk, with barycentric interpolation of vertex features
- A unified framework supporting novel view synthesis, scene reconstruction, and 3D appearance/geometry editing
- Image fitting benchmarks show higher PSNR with fewer parameters than hash encoding on all four test images (e.g. 59.81 dB vs 50.31 dB with 3.61M vs 5.75M params on Albert)
More from Multimodal
- Midjourney sref codes render knit and embroidered Adidas-style retro sportswear — sergeantsref · 2026-11-16
- Open-source ComfyUI node enhances MiniMax H3 image conditioning for better texture and detail — Away_Exam_4586 · 2026-11-16
- Combining LTX Audio Nodes with MiniMax H3 for Lip-Synced Video in ComfyUI — CQDSN · 2026-11-16
- One character, two styles: AI generates matching 2D illustration and 3D render — azed_ai · 2026-11-16
- Dev shares a skill that turns your project into a retro CD-install-disc cover — karminski3 · 2026-11-16
- Surreal AI images: three impossible worlds built with MiniMax's H3-Image — Hailuo_AI · 2026-11-16