Bridging the Gap to Real-World Object-Centric Learning
Code Website
Humans naturally decompose their environment into entities at the appropriate level of abstraction to act in the world. Allowing machine learning algorithms to derive this decomposition in an unsupervised way has become an important line of research. However, current methods are restricted to simulated data or require additional information in the form of motion or depth in order to successfully discover objects. In this work, we overcome this limitation by showing that reconstructing features from models trained in a self-supervised manner is a sufficient training signal for object-centric representations to arise in a fully unsupervised way. Our approach, DINOSAUR, significantly out-performs existing object-centric learning models on simulated data and is the first unsupervised object-centric model that scales to real world-datasets such as COCO and PASCAL VOC. DINOSAUR is conceptually simple and shows competitive performance compared to more involved pipelines from the computer vision literature.
| Author(s): | Maximilian Seitzer and Max Horn and Andrii Zadaianchuk and Dominik Zietlow and Tianjun Xiao and Carl-Johann Simon-Gabriel and Tong He and Zheng Zhang and Bernhard Schölkopf and Thomas Brox and Francesco Locatello |
| Links: | |
| Book Title: | Proceedings of the Eleventh International Conference on Learning Representations |
| Year: | 2023 |
| Month: | May |
| Bibtex Type: | Conference Paper (inproceedings) |
| Event Name: | The Eleventh International Conference on Learning Representations (ICLR) |
| Event Place: | Rwanda, Africa |
| State: | Published |
| URL: | https://openreview.net/forum?id=b9tUk-f_aG |
| Electronic Archiving: | grant_archive |
BibTex
@inproceedings{Seitzer2023BridgingTheGap,
title = {Bridging the Gap to Real-World Object-Centric Learning},
booktitle = {Proceedings of the Eleventh International Conference on Learning Representations},
abstract = {Humans naturally decompose their environment into entities at the appropriate level of abstraction to act in the world. Allowing machine learning algorithms to derive this decomposition in an unsupervised way has become an important line of research. However, current methods are restricted to simulated data or require additional information in the form of motion or depth in order to successfully discover objects. In this work, we overcome this limitation by showing that reconstructing features from models trained in a self-supervised manner is a sufficient training signal for object-centric representations to arise in a fully unsupervised way. Our approach, DINOSAUR, significantly out-performs existing object-centric learning models on simulated data and is the first unsupervised object-centric model that scales to real world-datasets such as COCO and PASCAL VOC. DINOSAUR is conceptually simple and shows competitive performance compared to more involved pipelines from the computer vision literature. },
month = may,
year = {2023},
slug = {seitzer2023bridgingthegap},
author = {Seitzer, Maximilian and Horn, Max and Zadaianchuk, Andrii and Zietlow, Dominik and Xiao, Tianjun and Simon-Gabriel, Carl-Johann and He, Tong and Zhang, Zheng and Sch{\"o}lkopf, Bernhard and Brox, Thomas and Locatello, Francesco},
url = {https://openreview.net/forum?id=b9tUk-f_aG},
month_numeric = {5}
}