MetricGrids: Arbitrary Nonlinear Approximation with Elementary Metric Grids based Implicit Neural Representation
Shu Wang, Yanbo Gao, Shuai Li, Chong Lv, Xun Cai, Chuankun Li, Hui Yuan, Jinglin Zhang · 2025
Read MetricGrids: Arbitrary Nonlinear Approximation with Elementary Metric Grids based Implicit Neural Representation by Shu Wang, Yanbo Gao, Shuai Li, Chong Lv, Xun Cai, Chuankun Li, Hui Yuan, Jinglin Zhang (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-10000.txt.
Description
Shu Wang and seven co-authors present MetricGrids, a new way for neural networks to store signals, in a 2025 paper. Grid-based storage is fast and compact for images and 3D scenes. But plain grids store features along straight scales, which can miss curves and fine bends in complex signals.
MetricGrids adds extra grids measured in different distance spaces, acting like higher-order terms in a classic expansion. A hashing step keeps storage small while limiting clashes between stored values. A smart decoder then reads the grids without needing huge memory. Tests on image and 3D tasks showed sharper fits and better rendering than rival grid methods. The design targets high-quality scenes that still fit in modest storage.
Catalog record
| CREATOR | Shu Wang, Yanbo Gao, Shuai Li, Chong Lv, Xun Cai, Chuankun Li, Hui Yuan, Jinglin Zhang |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 38 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-10000