Temporally Consistent Object-Centric Learning by Contrasting Slots

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Manasyan, Anna, Seitzer, Maximilian, Radovic, Filip, Martius, Georg, Zadaianchuk, Andrii
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916655791079424
author Manasyan, Anna
Seitzer, Maximilian
Radovic, Filip
Martius, Georg
Zadaianchuk, Andrii
author_facet Manasyan, Anna
Seitzer, Maximilian
Radovic, Filip
Martius, Georg
Zadaianchuk, Andrii
contents Unsupervised object-centric learning from videos is a promising approach to extract structured representations from large, unlabeled collections of videos. To support downstream tasks like autonomous control, these representations must be both compositional and temporally consistent. Existing approaches based on recurrent processing often lack long-term stability across frames because their training objective does not enforce temporal consistency. In this work, we introduce a novel object-level temporal contrastive loss for video object-centric models that explicitly promotes temporal consistency. Our method significantly improves the temporal consistency of the learned object-centric representations, yielding more reliable video decompositions that facilitate challenging downstream tasks such as unsupervised object dynamics prediction. Furthermore, the inductive bias added by our loss strongly improves object discovery, leading to state-of-the-art results on both synthetic and real-world datasets, outperforming even weakly-supervised methods that leverage motion masks as additional cues.
format Preprint
id arxiv_https___arxiv_org_abs_2412_14295
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Temporally Consistent Object-Centric Learning by Contrasting Slots
Manasyan, Anna
Seitzer, Maximilian
Radovic, Filip
Martius, Georg
Zadaianchuk, Andrii
Computer Vision and Pattern Recognition
Artificial Intelligence
Machine Learning
Robotics
Unsupervised object-centric learning from videos is a promising approach to extract structured representations from large, unlabeled collections of videos. To support downstream tasks like autonomous control, these representations must be both compositional and temporally consistent. Existing approaches based on recurrent processing often lack long-term stability across frames because their training objective does not enforce temporal consistency. In this work, we introduce a novel object-level temporal contrastive loss for video object-centric models that explicitly promotes temporal consistency. Our method significantly improves the temporal consistency of the learned object-centric representations, yielding more reliable video decompositions that facilitate challenging downstream tasks such as unsupervised object dynamics prediction. Furthermore, the inductive bias added by our loss strongly improves object discovery, leading to state-of-the-art results on both synthetic and real-world datasets, outperforming even weakly-supervised methods that leverage motion masks as additional cues.
title Temporally Consistent Object-Centric Learning by Contrasting Slots
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Machine Learning
Robotics
url https://arxiv.org/abs/2412.14295