Research map / Detection and segmentation / CNN features
Learning Instance Occlusion for 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 |
|---|---|---|---|
| Add a learned instance occlusion head to Panoptic FPN fusion | Panoptic FPN with ResNet-50 · COCO val | PQ: 39.5 -> 41.3; +1.8 | Helped |
| Add a learned instance occlusion head to Panoptic FPN fusion | Panoptic FPN with ResNet-101 · COCO val | PQ: 41.0 -> 43.0; +2.0 | Helped |
| Add a learned instance occlusion head to Panoptic FPN fusion | Panoptic FPN with ResNet-50 · Cityscapes val | PQ: 58.6 -> 59.3; +0.7 | Helped |
| Enable multi-scale testing | OCFusion* with ResNet-50 · Cityscapes val | PQ: 59.3 -> 60.2 | Helped |
| Enable inter-class occlusion handling (ablation) | Panoptic FPN with ResNet-50 · Cityscapes val | PQ: 58.6 -> 59.2 (+0.5) | Helped |
| Add intra-class occlusion handling to inter-class occlusion handling (ablation) | OCFusion with ResNet-50 · Cityscapes val | PQ: 59.2 -> 59.3 (+0.7) | Helped |
| Set τ=0.4 and ρ=0.05 (ablation) | OCFusion · COCO val | PQ: 41.27 | Helped |
| Raise τ from 0.4 to 0.6 at ρ=0.10 (ablation) | OCFusion · COCO val | PQ: 41.22 -> 41.03 | Hurt |
| Set τ=0.6 and ρ=0.10 (ablation) | OCFusion · Cityscapes val | PQ: 59.33 | Helped |
| Raise ρ from 0.10 to 0.20 at τ=0.6 (ablation) | OCFusion · Cityscapes val | PQ: 59.33 -> 58.70 | Hurt |
| Add the occlusion head | Panoptic FPN · COCO | inference runtime: 153 -> 156; +3 | Hurt |
| Add the occlusion head | Panoptic FPN · Cityscapes | inference runtime: 378 -> 396; +18 | Hurt |
Closest papers by meaning
- BANet: Bidirectional Aggregation Network With Occlusion Handling for Panoptic Segmentation (CVPR 2020 · similarity 0.90)
- An End-To-End Network for Panoptic Segmentation (CVPR 2019 · similarity 0.90)
- Robust Instance Segmentation Through Reasoning About Multi-Object Occlusion (CVPR 2021 · similarity 0.89)
- Deep Occlusion-Aware Instance Segmentation With Overlapping BiLayers (CVPR 2021 · similarity 0.88)
- PanoOcc: Unified Occupancy Representation for Camera-based 3D Panoptic Segmentation (CVPR 2024 · similarity 0.85)
- Revisiting Open-Set Panoptic Segmentation (AAAI 2024 · 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.