Research map / Detection and segmentation / Open-set and few-shot detection
Simple Copy-Paste Is a Strong Data Augmentation Method for Instance 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 random object Copy-Paste augmentation | Cascade Eff-B7 NAS-FPN · COCO test-dev | Box AP: 54.8 -> 56.0 | Helped |
| Combine Copy-Paste with self-training and pseudo-labeled data | Cascade Eff-B7 NAS-FPN · COCO test-dev | Mask AP: 46.9 -> 49.1 | Helped |
| Add Copy-Paste augmentation | Eff-B7 FPN (640) · LVIS v1.0 val, single-stage | Mask AP_r: 9.7 -> 12.8 | Hurt |
| Combine RFS with Copy-Paste | Eff-B7 FPN (640) · LVIS v1.0 val, single-stage | Mask AP_r: 9.7 -> 18.4 | Helped |
| Add Copy-Paste augmentation | Eff-B7 FPN (640) · LVIS v1.0 val, two-stage | Mask AP: 31.3 -> 33.0 | Helped |
| Add RFS | Eff-B7 FPN (640), two-stage LVIS training · LVIS v1.0 val, two-stage | Mask AP: 31.3 -> 30.1 | Hurt |
| Paste one random object instead of a random subset (ablation) | Copy-Paste on EfficientNetB7-FPN (640) · COCO | Box AP: 49.1 vs 50.0 | Hurt |
| Paste all objects instead of a random subset (ablation) | Copy-Paste on EfficientNetB7-FPN (640) · COCO | Box AP: 49.7 vs 50.0 | Hurt |
| Remove alpha blending (ablation) | Copy-Paste on EfficientNetB7-FPN (640) · COCO | Box AP: 50.0 -> 50.0 | No clear effect |
| Use standard scale jittering for both the main and pasted images (ablation) | Copy-Paste on EfficientNetB7-FPN (640) · COCO | Box AP: 48.1 vs 50.0 | Hurt |
| Use SSJ for the pasted image instead of LSJ while retaining LSJ for the main image (ablation) | Copy-Paste on EfficientNetB7-FPN (640) · COCO | Box AP: 49.7 vs 50.0 | Hurt |
| Paste ground-truth objects into both COCO and pseudo-labeled images rather than COCO alone (ablation) | Self-training with Copy-Paste · COCO | Box AP: 50.4 -> 51.4 | Helped |
Closest papers by meaning
- X-Paste: Revisiting Scalable Copy-Paste for Instance Segmentation using CLIP and StableDiffusion (ICML 2023 · similarity 0.94)
- FAPIS: A Few-Shot Anchor-Free Part-Based Instance Segmenter (CVPR 2021 · similarity 0.87)
- The Devil is in the Object Boundary: Towards Annotation-free Instance Segmentation using Foundation Models (ICLR 2024 · similarity 0.86)
- Training-Free Dataset Pruning for Instance Segmentation (ICLR 2025 · similarity 0.86)
- MaskDiff: Modeling Mask Distribution with Diffusion Probabilistic Model for Few-Shot Instance Segmentation (AAAI 2024 · similarity 0.86)
- WISH: Weakly Supervised Instance Segmentation using Heterogeneous Labels (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.