Research map / Detection and segmentation / Open-set and few-shot detection
Amodal Instance Segmentation with IRAIS Dataset for Sim-to-Real Transfer
Bidong Chen, Lingui Li
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 |
|---|---|---|---|
| use unified query set instead of dual-set queries (ablation) | EUREKA · 3D-IRAIS | amodal AP: 88.88 -> 89.10 | Improved |
| use unified query set instead of twin queries (ablation) | EUREKA · 3D-IRAIS | amodal AP: 88.49 -> 89.10 | Improved |
| use unified query set instead of twin queries (ablation) | EUREKA · 3D-IRAIS | GFLOPs: 846 -> 690 | Improved |
| use DINOv3 instead of DINOv2 backbone | EUREKA · 3D-IRAIS | amodal AP: 89.10 -> 89.75 | Improved |
| use DINOv3 instead of DINOv2 backbone | EUREKA · Real-IRAIS (sim-to-real) | amodal AP: 37.10 -> 33.00 | Hurt |
| use DINOv3 instead of DINOv2 backbone | EUREKA · COCOA-cls | amodal AP: 37.82 -> 39.50 | Improved |
| use DINOv3 instead of DINOv2 backbone | EUREKA · D2SA | amodal AP: 82.35 -> 84.84 | Improved |
| choose union versus visible or amodal mask conditioning in attention (ablation) | EUREKA · 3D-IRAIS vs COCOA-cls | AP | Mixed |
| evaluate on 5x larger Real-IRAIS-Aug instead of Real-IRAIS | EUREKA DINOv2 · Real-IRAIS | amodal AP: 37.10 -> 34.87 | No clear effect |
| evaluate on 5x larger Real-IRAIS-Aug instead of Real-IRAIS | EUREKA DINOv3 · Real-IRAIS | amodal AP: 33.00 -> 32.00 | No clear effect |
| replace AISFormer with EUREKA for sim-to-real transfer | AISFormer ResNet-101 · Real-IRAIS | amodal AP: 18.82 -> 37.10 | Improved |
| replace AISFormer+SAMEO with EUREKA-DINOv3 | AISFormer+SAMEO · D2SA | amodal AP: 74.1 -> 84.84 | Improved |
Closest papers by meaning
- BLADE: Box-Level Supervised Amodal Segmentation through Directed Expansion (AAAI 2024 · similarity 0.86)
- Amodal Scene Analysis via Holistic Occlusion Relation Inference and Generative Mask Completion (AAAI 2024 · similarity 0.86)
- Segment Anything, Even Occluded (CVPR 2025 · similarity 0.86)
- Amodal Ground Truth and Completion in the Wild (CVPR 2024 · similarity 0.85)
- pix2gestalt: Amodal Segmentation by Synthesizing Wholes (CVPR 2024 · similarity 0.85)
- Towards Efficient Foundation Model for Zero-shot Amodal Segmentation (CVPR 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.
