3D Shape Completion with Test-Time Training
Authored by Michael Schopf-Kuester, Zorah Lähner, Michael Moeller
Published in ICML Workshop on Geometry-Grounded Representation Learning and Generative Modeling (GRaM) 2024
Abstract
This work addresses the problem of shape completion, i.e., the task of restoring incomplete shapes by predicting their missing parts. While several prior works have focused on directly predicting the entire restored shape at once, we describe the prediction of the fractured and newly restored parts in a separate (but coupled) manner. We use a decoder network motivated by related work on the prediction of signed distance functions (DeepSDF). In particular, our representation allows us to consider test-time-training, i.e., finetuning network parameters to match the given incomplete shape more accurately during inference. While previous works often have difficulties with artifacts around the fracture boundary, we demonstrate that our overfitting to the fractured parts leads to significant improvements in the restoration of eight different shape categories of the ShapeNet data set in terms of their chamfer distances.
Resources
Bibtex
@inproceedings{ schopf2024completion,
author = { Michael Schopf-Kuester and Zorah Lähner and Michael Moeller },
title = { 3D Shape Completion with Test-Time Training },
booktitle = { ICML Workshop on Geometry-Grounded Representation Learning and Generative Modeling (GRaM) },
year = { 2024 },
}