Overfit Neural Networks as a Compact Shape Representation - Jianfei Guo

目录

目录

Overfit Neural Networks as a Compact Shape Representation

Jianfei Guo 出版于 paper_reading

2020-12-03 2020-12-03 约 162 字预计阅读 1 分钟

<OverfitSDF> Overfit neural networks as a compact shape representation

arxiv2020 arXiv e-prints

Thomas Davies, Derek Nowrouzezahrai, Alec Jacobson

University of Toronto, McGill University

compact representation

目录

Motivation

现在的DeepSDF倾向于做category类别的泛化/生成；
本篇主要提出其实overfit到一个具体的shape的SDF可以作为mesh的一种更compact紧致的表征，而且相比于显式地mesh更省空间
同时，做了很多具体shape optimization的优化，比如采样时基于重要度采样，一些biased points，等