MUFASA: A Multi-Layer Framework for Slot Attention
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866911431062978560 |
|---|---|
| author | Bock, Sebastian Schüßler, Leonie Singh, Krishnakant Schaub-Meyer, Simone Roth, Stefan |
| author_facet | Bock, Sebastian Schüßler, Leonie Singh, Krishnakant Schaub-Meyer, Simone Roth, Stefan |
| contents | Unsupervised object-centric learning (OCL) decomposes visual scenes into distinct entities. Slot attention is a popular approach that represents individual objects as latent vectors, called slots. Current methods obtain these slot representations solely from the last layer of a pre-trained vision transformer (ViT), ignoring valuable, semantically rich information encoded across the other layers. To better utilize this latent semantic information, we introduce MUFASA, a lightweight plug-and-play framework for slot attention-based approaches to unsupervised object segmentation. Our model computes slot attention across multiple feature layers of the ViT encoder, fully leveraging their semantic richness. We propose a fusion strategy to aggregate slots obtained on multiple layers into a unified object-centric representation. Integrating MUFASA into existing OCL methods improves their segmentation results across multiple datasets, setting a new state of the art while simultaneously improving training convergence with only minor inference overhead. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_07544 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | MUFASA: A Multi-Layer Framework for Slot Attention Bock, Sebastian Schüßler, Leonie Singh, Krishnakant Schaub-Meyer, Simone Roth, Stefan Computer Vision and Pattern Recognition Unsupervised object-centric learning (OCL) decomposes visual scenes into distinct entities. Slot attention is a popular approach that represents individual objects as latent vectors, called slots. Current methods obtain these slot representations solely from the last layer of a pre-trained vision transformer (ViT), ignoring valuable, semantically rich information encoded across the other layers. To better utilize this latent semantic information, we introduce MUFASA, a lightweight plug-and-play framework for slot attention-based approaches to unsupervised object segmentation. Our model computes slot attention across multiple feature layers of the ViT encoder, fully leveraging their semantic richness. We propose a fusion strategy to aggregate slots obtained on multiple layers into a unified object-centric representation. Integrating MUFASA into existing OCL methods improves their segmentation results across multiple datasets, setting a new state of the art while simultaneously improving training convergence with only minor inference overhead. |
| title | MUFASA: A Multi-Layer Framework for Slot Attention |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2602.07544 |