Research map / Detection and segmentation / Open-set and few-shot detection
Conditional Object-Centric Learning from Video
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use sequential SAVi instead of the strongest listed unsupervised baseline | CATER · CATER | FG-ARI: 91.8 -> 92.8 | Helped |
| Add first-frame bounding-box conditioning and optical-flow prediction | SAVi · MOVi++ | FG-ARI: 47.6 -> 77.4 | Helped |
| Replace the Transformer predictor with the identity function (ablation) | SAVi + Bounding Box · MOVi++ | FG-ARI: 82.0 -> 83.0 | Helped |
| Replace the Transformer predictor with the identity function (ablation) | SAVi + Bounding Box · MOVi++ | mIoU: 54.3 -> 44.7 | Hurt |
| Replace the Transformer predictor with a slot-wise shared MLP (ablation) | SAVi + Bounding Box · MOVi++ | mIoU: 54.3 -> 49.5 | Hurt |
| Normalize corrector attention over visual features rather than slots (ablation) | SAVi + Bounding Box · MOVi++ | mIoU: 54.3 -> 11.4 | Hurt |
| Use estimated optical flow instead of ground-truth flow (ablation) | SAVi · MOVi and MOVi++ | segmentation quality | No clear effect |
| Replace optical-flow prediction with RGB reconstruction (ablation) | SAVi + Bounding Box · MOVi++ | segmentation quality | Hurt |
| Replace optical-flow prediction with RGB reconstruction (ablation) | SAVi + Bounding Box · MOVi | segmentation quality | No clear effect |
| Remove conditional hints and use learned fixed initial slots (ablation) | SAVi · MOVi++ | FG-ARI: 77.4 -> 47.6 | Hurt |
| Use a ResNet encoder (ablation) | SAVi + Bounding Box · MOVi++ | mIoU: 45.9 -> 50.7 | Helped |
| Use bounding boxes as matching-based training supervision rather than as conditional slot initialization (ablation) | SAVi · MOVi | FG-ARI: 92.9 -> 83.0; 92.9 -> 82.3 | Hurt |
Closest papers by meaning
- Shepherding Slots to Objects: Towards Stable and Robust Object-Centric Learning (CVPR 2023 · similarity 0.91)
- Temporally Consistent Object-Centric Learning by Contrasting Slots (CVPR 2025 · similarity 0.90)
- Segmenting Moving Objects via an Object-Centric Layered Representation (NeurIPS 2022 · similarity 0.90)
- Object-Centric Pretraining via Target Encoder Bootstrapping (ICLR 2025 · similarity 0.88)
- Using Diffusion Priors for Video Amodal Segmentation (CVPR 2025 · similarity 0.88)
- DORSal: Diffusion for Object-centric Representations of Scenes $\textit{et al.}$ (ICLR 2024 · similarity 0.87)
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.