Learning Gradient Fields for Shape Generation
Publikation: Bidrag til tidsskrift › Konferenceartikel › Forskning › fagfællebedømt
In this work, we propose a novel technique to generate shapes from point cloud data. A point cloud can be viewed as samples from a distribution of 3D points whose density is concentrated near the surface of the shape. Point cloud generation thus amounts to moving randomly sampled points to high-density areas. We generate point clouds by performing stochastic gradient ascent on an unnormalized probability density, thereby moving sampled points toward the high-likelihood regions. Our model directly predicts the gradient of the log density field and can be trained with a simple objective adapted from score-based generative models. We show that our method can reach state-of-the-art performance for point cloud auto-encoding and generation, while also allowing for extraction of a high-quality implicit surface. Code is available at https://github.com/RuojinCai/ShapeGF.
Originalsprog | Engelsk |
---|---|
Tidsskrift | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
Sider (fra-til) | 364-381 |
Antal sider | 18 |
ISSN | 0302-9743 |
DOI | |
Status | Udgivet - 2020 |
Eksternt udgivet | Ja |
Begivenhed | 16th European Conference on Computer Vision, ECCV 2020 - Glasgow, Storbritannien Varighed: 23 aug. 2020 → 28 aug. 2020 |
Konference
Konference | 16th European Conference on Computer Vision, ECCV 2020 |
---|---|
Land | Storbritannien |
By | Glasgow |
Periode | 23/08/2020 → 28/08/2020 |
Bibliografisk note
Publisher Copyright:
© 2020, Springer Nature Switzerland AG.
ID: 301818367