Neural Eulerian Scene Flow Fields

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Vedder, Kyle, Peri, Neehar, Khatri, Ishan, Li, Siyi, Eaton, Eric, Kocamaz, Mehmet, Wang, Yue, Yu, Zhiding, Ramanan, Deva, Pehserl, Joachim
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909943893852160
author Vedder, Kyle
Peri, Neehar
Khatri, Ishan
Li, Siyi
Eaton, Eric
Kocamaz, Mehmet
Wang, Yue
Yu, Zhiding
Ramanan, Deva
Pehserl, Joachim
author_facet Vedder, Kyle
Peri, Neehar
Khatri, Ishan
Li, Siyi
Eaton, Eric
Kocamaz, Mehmet
Wang, Yue
Yu, Zhiding
Ramanan, Deva
Pehserl, Joachim
contents We reframe scene flow as the task of estimating a continuous space-time ODE that describes motion for an entire observation sequence, represented with a neural prior. Our method, EulerFlow, optimizes this neural prior estimate against several multi-observation reconstruction objectives, enabling high quality scene flow estimation via pure self-supervision on real-world data. EulerFlow works out-of-the-box without tuning across multiple domains, including large-scale autonomous driving scenes and dynamic tabletop settings. Remarkably, EulerFlow produces high quality flow estimates on small, fast moving objects like birds and tennis balls, and exhibits emergent 3D point tracking behavior by solving its estimated ODE over long-time horizons. On the Argoverse 2 2024 Scene Flow Challenge, EulerFlow outperforms all prior art, surpassing the next-best unsupervised method by more than 2.5x, and even exceeding the next-best supervised method by over 10%.
format Preprint
id arxiv_https___arxiv_org_abs_2410_02031
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Neural Eulerian Scene Flow Fields
Vedder, Kyle
Peri, Neehar
Khatri, Ishan
Li, Siyi
Eaton, Eric
Kocamaz, Mehmet
Wang, Yue
Yu, Zhiding
Ramanan, Deva
Pehserl, Joachim
Computer Vision and Pattern Recognition
We reframe scene flow as the task of estimating a continuous space-time ODE that describes motion for an entire observation sequence, represented with a neural prior. Our method, EulerFlow, optimizes this neural prior estimate against several multi-observation reconstruction objectives, enabling high quality scene flow estimation via pure self-supervision on real-world data. EulerFlow works out-of-the-box without tuning across multiple domains, including large-scale autonomous driving scenes and dynamic tabletop settings. Remarkably, EulerFlow produces high quality flow estimates on small, fast moving objects like birds and tennis balls, and exhibits emergent 3D point tracking behavior by solving its estimated ODE over long-time horizons. On the Argoverse 2 2024 Scene Flow Challenge, EulerFlow outperforms all prior art, surpassing the next-best unsupervised method by more than 2.5x, and even exceeding the next-best supervised method by over 10%.
title Neural Eulerian Scene Flow Fields
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2410.02031