Research map / Detection and segmentation / Open-set and few-shot detection
MUFASA: A Multi-Layer Framework for Slot Attention
Sebastian Bock, Leonie Schüßler, Krishnakant Singh, Simone Schaub-Meyer, Stefan Roth
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add multi-layer slot attention with M-Fusion | SPOT · PASCAL VOC | mBO^c: 55.3 -> 59.8 | Improved |
| add multi-layer slot attention with M-Fusion | DINOSAUR · PASCAL VOC | mBO^c: 51.2 -> 57.6 | Improved |
| add multi-layer slot attention with M-Fusion | SPOT · MOVi-C | mBO^i: 47.0 -> 49.2 | Improved |
| add multi-layer slot attention with M-Fusion | DINOSAUR · MOVi-C | mBO^i: 42.4 -> 49.2 | Improved |
| add multi-layer slot attention with M-Fusion | SPOT · COCO | mIoU: 32.7 -> 32.5 | Hurt |
| add multi-layer slot attention with M-Fusion | SPOT · COCO | FG-ARI: 37.8 -> 35.6 | Hurt |
| add M-Fusion instead of Avg-Fusion (ablation) | SPOT-M · PASCAL VOC | mBO^c: 55.6 -> 59.8 | Improved |
| use MLP M-Fusion instead of T-Fusion (transformer layer) (ablation) | SPOT-M · PASCAL VOC | mBO^c: 59.0 -> 59.8 | Improved |
| use pairwise summation before fusion instead of concatenation (ablation) | SPOT-M · PASCAL VOC | mBO^c: 59.0 -> 59.8 | Improved |
| use last four consecutive layers instead of three or five (ablation) | SPOT-M · PASCAL VOC | segmentation metrics | Improved |
| combine earlier with later non-consecutive layers instead of last consecutive layers (ablation) | SPOT-M · PASCAL VOC | segmentation metrics | Hurt |
| add MUFASA with weaker MLP decoder | SPOT · COCO | mBO^c: 32.4 -> 34.7 | Improved |
Closest papers by meaning
- Adaptive Slot Attention: Object Discovery with Dynamic Slot Number (CVPR 2024 · similarity 0.89)
- Guided Slot Attention for Unsupervised Video Object Segmentation (CVPR 2024 · similarity 0.89)
- MetaSlot: Break Through the Fixed Number of Slots in Object-Centric Learning (NeurIPS 2025 · similarity 0.89)
- GLASS: Guided Latent Slot Diffusion for Object-Centric Learning (CVPR 2025 · similarity 0.89)
- SPOT: Self-Training with Patch-Order Permutation for Object-Centric Learning with Autoregressive Transformers (CVPR 2024 · similarity 0.89)
- Grounded Object-Centric Learning (ICLR 2024 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
