Research map / Video / Video segmentation and tracking
Selective Synergistic Learning for Video Object-Centric Learning
WonJun Moon, Jae-Pil Heo
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 boundary supervision L_bd (ablation) | SSync baseline · MOVi-C | FG-ARI: 69.0 -> 72.9 | Improved |
| add interior supervision L_nbd (ablation) | SSync baseline · MOVi-C | FG-ARI: 69.0 -> 71.4 | Improved |
| add L_nbd on top of L_bd (ablation) | SSync · MOVi-C | FG-ARI: 72.9 -> 77.1 | Improved |
| add transitive pseudo-label merging (ablation) | SSync · MOVi-C | FG-ARI: 77.1 -> 79.4 | Improved |
| use transitive merging instead of pairwise merging (ablation) | SSync · MOVi-C | FG-ARI: 78.2 -> 79.4 | Improved |
| use transitive merging instead of slot regularization (ablation) | SSync · MOVi-C | FG-ARI: 74.6 -> 79.4 | Improved |
| increase slot count from 11 to 15 (ablation) | SlotContrast · MOVi-C | FG-ARI: 69.3 -> 61.8 | Hurt |
| increase slot count from 7 to 15 (ablation) | SSync · MOVi-C | FG-ARI: 76.9 -> 78.8 | No clear effect |
| use encoder boundaries and decoder interiors instead of the reverse (ablation) | SSync · MOVi-C | FG-ARI: 67.6 -> 77.1 | Improved |
| use hard pseudo-labels instead of soft targets (ablation) | SSync · MOVi-C | FG-ARI: 72.9 -> 77.1 | Improved |
| use local-consistency selection instead of erosion (ablation) | SSync · MOVi-C | FG-ARI: 76.4 -> 77.1 | Improved |
| vary SSync warmup ratio from 30% to 5% (ablation) | SSync · MOVi-C | FG-ARI: 79.4 -> 79.1 | No clear effect |
Closest papers by meaning
- Temporally Consistent Object-Centric Learning by Contrasting Slots (CVPR 2025 · similarity 0.87)
- Object-Centric Pretraining via Target Encoder Bootstrapping (ICLR 2025 · similarity 0.86)
- Parallelized Spatiotemporal Slot Binding for Videos (ICML 2024 · similarity 0.86)
- Bootstrapping Top-down Information for Self-modulating Slot Attention (NeurIPS 2024 · similarity 0.85)
- SPOT: Self-Training with Patch-Order Permutation for Object-Centric Learning with Autoregressive Transformers (CVPR 2024 · similarity 0.85)
- MetaSlot: Break Through the Fixed Number of Slots in Object-Centric Learning (NeurIPS 2025 · similarity 0.85)
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.
