Research map / Detection and segmentation / CNN features
You Only Segment Once: Towards Real-Time Panoptic Segmentation
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 convolution-first aggregation (CFA) instead of interpolation-first (IFA) (ablation) | YOSO · COCO val | latency (us): 4871 -> 1877 | Helped |
| Use CFA instead of IFA (ablation) | YOSO · COCO val | PQ: 47.5 -> 47.0 | Hurt |
| Use separable dynamic convolution attention (SDCA) instead of multi-head cross-attention (ablation) | YOSO · COCO val | PQ: 46.0 -> 47.0 | Helped |
| Use SDCA instead of MHCA (ablation) | YOSO · COCO val | latency (us): 2608 -> 2183 | Helped |
| Add cross-dimension interaction (depthwise) via full SDCA instead of pointwise-only PDCA (ablation) | YOSO · COCO val | PQ: 43.7 -> 47.0 | Helped |
| Add pointwise cross-token interaction to depthwise-only DDCA (full SDCA) (ablation) | YOSO · COCO val | PQ: 46.6 -> 47.0 | Helped |
| Increase attention blocks from 1 to 3 (ablation) | YOSO · COCO val | PQ: 46.4 -> 47.5 | Mixed |
| Increase iteration stages from 1 to 3 (ablation) | YOSO · COCO val | PQ: 45.7 -> 47.5 | Mixed |
| Increase proposal kernels from 50 to 100 (ablation) | YOSO · COCO val | PQ: 44.2 -> 47.0 | Helped |
| Increase proposal kernels from 100 to 200 (ablation) | YOSO · COCO val | PQ: 47.0 -> 46.7 | Hurt |
| Increase dynamic conv kernel size from 1 to 5 (ablation) | YOSO · COCO val | PQ: 46.8 -> 47.3 | Helped |
| Increase kernel size from 5 to 7 (ablation) | YOSO · COCO val | PQ: 47.3 -> 47.1 | No clear effect |
Closest papers by meaning
- Real-Time Panoptic Segmentation From Dense Detections (CVPR 2020 · similarity 0.91)
- Panoptic-DeepLab: A Simple, Strong, and Fast Baseline for Bottom-Up Panoptic Segmentation (CVPR 2020 · similarity 0.90)
- Seamless Scene Segmentation (CVPR 2019 · similarity 0.89)
- Auto-Panoptic: Cooperative Multi-Component Architecture Search for Panoptic Segmentation (NeurIPS 2020 · similarity 0.89)
- Fully Convolutional Networks for Panoptic Segmentation (CVPR 2021 · similarity 0.88)
- Scene-Centric Unsupervised Panoptic Segmentation (CVPR 2025 · 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.