Research map / Detection and segmentation / Open-set and few-shot detection
Slot-VPS: Object-Centric Representation Learning for Video Panoptic Segmentation
Check 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 |
|---|---|---|---|
| Represent video panoptic entities with panoptic slots and process them using the unified end-to-end framework | Slot-VPS · Cityscapes-VPS val | VPQ: 57.3 -> 59.7 | Helped |
| Represent video panoptic entities with panoptic slots and process them using the unified end-to-end framework | Slot-VPS · Cityscapes-VPS test | VPQ: 57.8 -> 60.1 | Helped |
| Use Slot-VPS with Swin-L-FPN instead of ViP-DeepLab with WR-41 | Cityscapes-VPS · Cityscapes-VPS val | VPQ: 63.1 -> 63.7 | Helped |
| Use Slot-VPS with Swin-L-FPN instead of ViP-DeepLab with WR-41 | Cityscapes-VPS · Cityscapes-VPS test | VPQ: 62.5 -> 63.3 | Helped |
| Use Slot-VPS with ResNet50-FPN instead of VPSNet with ResNet50-FPN | VIPER · VIPER val | VPQ: 51.9 -> 53.2 | Helped |
| Use slot-dimension softmax rather than spatial-dimension softmax (ablation) | Slot-VPS Retriever · Cityscapes-VPS val | PQ: 53.5 -> 60.5 | Helped |
| Add a final projection layer after attention-weighted value retrieval (ablation) | Slot-VPS Retriever · Cityscapes-VPS val | PQ: 60.5 -> 60.7 | Helped |
| Use multi-scale features rather than a single scale (ablation) | Slot-VPS · Cityscapes-VPS val | PQ: 58.9 -> 60.5 | Helped |
| Set the slot count to 100 rather than 50, 80, 200, or 300 (ablation) | Slot-VPS · Cityscapes-VPS val | PQ: 50 slots: 57.4; 80 slots: 57.6; 100 slots: 60.5; 200 slots: 58.7; 300 slots: 58.2 | Helped |
| Increase the number of VPR modules per scale (ablation) | Slot-VPS · Cityscapes-VPS val | PQ: 56.4 -> 58.9 -> 60.1 | Helped |
| Use VPR module counts 1/2/2/2 instead of 1/1/1/1 (ablation) | Slot-VPS with FFN hidden dimension 2048 · Cityscapes-VPS val | PQ: 51.6 -> 60.5 | Helped |
| Remove the Video Retriever (ablation) | Slot-VPS · Cityscapes-VPS val | VPQ: 1.5 VPQ drop | Hurt |
Closest papers by meaning
- Video Panoptic Segmentation (CVPR 2020 · similarity 0.92)
- Scene-Centric Unsupervised Video Panoptic Segmentation (CVPR 2026 · similarity 0.90)
- Hierarchical Lovasz Embeddings for Proposal-Free Panoptic Segmentation (CVPR 2021 · similarity 0.88)
- Scene-Centric Unsupervised Panoptic Segmentation (CVPR 2025 · similarity 0.87)
- Task-aligned Part-aware Panoptic Segmentation through Joint Object-Part Representations (CVPR 2024 · similarity 0.87)
- Guided Slot Attention for Unsupervised Video Object Segmentation (CVPR 2024 · similarity 0.86)
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.