Boosting Object Representation Learning via Motion and Object Continuity
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2022
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866916132370251776 |
|---|---|
| author | Delfosse, Quentin Stammer, Wolfgang Rothenbacher, Thomas Vittal, Dwarak Kersting, Kristian |
| author_facet | Delfosse, Quentin Stammer, Wolfgang Rothenbacher, Thomas Vittal, Dwarak Kersting, Kristian |
| contents | Recent unsupervised multi-object detection models have shown impressive performance improvements, largely attributed to novel architectural inductive biases. Unfortunately, they may produce suboptimal object encodings for downstream tasks. To overcome this, we propose to exploit object motion and continuity, i.e., objects do not pop in and out of existence. This is accomplished through two mechanisms: (i) providing priors on the location of objects through integration of optical flow, and (ii) a contrastive object continuity loss across consecutive image frames. Rather than developing an explicit deep architecture, the resulting Motion and Object Continuity (MOC) scheme can be instantiated using any baseline object detection model. Our results show large improvements in the performances of a SOTA model in terms of object discovery, convergence speed and overall latent object representations, particularly for playing Atari games. Overall, we show clear benefits of integrating motion and object continuity for downstream tasks, moving beyond object representation learning based only on reconstruction. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2211_09771 |
| institution | arXiv |
| publishDate | 2022 |
| record_format | arxiv |
| spellingShingle | Boosting Object Representation Learning via Motion and Object Continuity Delfosse, Quentin Stammer, Wolfgang Rothenbacher, Thomas Vittal, Dwarak Kersting, Kristian Computer Vision and Pattern Recognition Machine Learning Recent unsupervised multi-object detection models have shown impressive performance improvements, largely attributed to novel architectural inductive biases. Unfortunately, they may produce suboptimal object encodings for downstream tasks. To overcome this, we propose to exploit object motion and continuity, i.e., objects do not pop in and out of existence. This is accomplished through two mechanisms: (i) providing priors on the location of objects through integration of optical flow, and (ii) a contrastive object continuity loss across consecutive image frames. Rather than developing an explicit deep architecture, the resulting Motion and Object Continuity (MOC) scheme can be instantiated using any baseline object detection model. Our results show large improvements in the performances of a SOTA model in terms of object discovery, convergence speed and overall latent object representations, particularly for playing Atari games. Overall, we show clear benefits of integrating motion and object continuity for downstream tasks, moving beyond object representation learning based only on reconstruction. |
| title | Boosting Object Representation Learning via Motion and Object Continuity |
| topic | Computer Vision and Pattern Recognition Machine Learning |
| url | https://arxiv.org/abs/2211.09771 |