Self-Supervised Learning of Motion Concepts by Optimizing Counterfactuals

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Stojanov, Stefan, Wendt, David, Kim, Seungwoo, Venkatesh, Rahul, Feigelis, Kevin, Wu, Jiajun, Yamins, Daniel LK
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909552304193536
author Stojanov, Stefan
Wendt, David
Kim, Seungwoo
Venkatesh, Rahul
Feigelis, Kevin
Wu, Jiajun
Yamins, Daniel LK
author_facet Stojanov, Stefan
Wendt, David
Kim, Seungwoo
Venkatesh, Rahul
Feigelis, Kevin
Wu, Jiajun
Yamins, Daniel LK
contents Estimating motion in videos is an essential computer vision problem with many downstream applications, including controllable video generation and robotics. Current solutions are primarily trained using synthetic data or require tuning of situation-specific heuristics, which inherently limits these models' capabilities in real-world contexts. Despite recent developments in large-scale self-supervised learning from videos, leveraging such representations for motion estimation remains relatively underexplored. In this work, we develop Opt-CWM, a self-supervised technique for flow and occlusion estimation from a pre-trained next-frame prediction model. Opt-CWM works by learning to optimize counterfactual probes that extract motion information from a base video model, avoiding the need for fixed heuristics while training on unrestricted video inputs. We achieve state-of-the-art performance for motion estimation on real-world videos while requiring no labeled data.
format Preprint
id arxiv_https___arxiv_org_abs_2503_19953
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Self-Supervised Learning of Motion Concepts by Optimizing Counterfactuals
Stojanov, Stefan
Wendt, David
Kim, Seungwoo
Venkatesh, Rahul
Feigelis, Kevin
Wu, Jiajun
Yamins, Daniel LK
Computer Vision and Pattern Recognition
Estimating motion in videos is an essential computer vision problem with many downstream applications, including controllable video generation and robotics. Current solutions are primarily trained using synthetic data or require tuning of situation-specific heuristics, which inherently limits these models' capabilities in real-world contexts. Despite recent developments in large-scale self-supervised learning from videos, leveraging such representations for motion estimation remains relatively underexplored. In this work, we develop Opt-CWM, a self-supervised technique for flow and occlusion estimation from a pre-trained next-frame prediction model. Opt-CWM works by learning to optimize counterfactual probes that extract motion information from a base video model, avoiding the need for fixed heuristics while training on unrestricted video inputs. We achieve state-of-the-art performance for motion estimation on real-world videos while requiring no labeled data.
title Self-Supervised Learning of Motion Concepts by Optimizing Counterfactuals
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2503.19953