Research map / Detection and segmentation / Open-set and few-shot detection
FAPIS: A Few-Shot Anchor-Free Part-Based Instance Segmenter
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 |
|---|---|---|---|
| Use FAPIS instead of the strongest baseline | one-shot instance segmentation · COCO-20^i | mAP50: 19.0 vs 17.2 | Helped |
| Use FAPIS instead of the strongest baseline | five-shot instance segmentation · COCO-20^i | mAP50: 21.2 vs 19.2 | Helped |
| Use anchor-free detection instead of anchor-based RPN (ablation) | FAPIS · COCO-20^0, K=1 | object detection mAP50: 18.4 -> 20.9 | Helped |
| Use anchor-free detection instead of anchor-based RPN (ablation) | FAPIS · COCO-20^0, K=1 | instance segmentation mAP50: 16.4 -> 18.0 | Helped |
| Add the part-based instance segmentor instead of the Mask-RCNN mask head (ablation) | FAIS · COCO-20^0, K=1 | instance segmentation mAP50: 18.0 -> 20.9 | Helped |
| Add SimNet (ablation) | FAIS · COCO-20^0, K=1 | object detection mAP50: 20.2 -> 20.9 | Helped |
| Add channel-wise multiplication alongside SimNet (ablation) | FAIS-CWM · COCO-20^0, K=1 | object detection mAP50: 16.5 -> 20.9 | Helped |
| Use the part-based segmentor instead of an RPN-based detector with the Mask-RCNN head (ablation) | FIS · COCO-20^0, K=1 | instance segmentation mAP50: 16.4 -> 17.4 | Helped |
| Add NMF regularization (ablation) | FAPIS · COCO-20^0, K=1 | instance segmentation mAP50: 18.4 -> 18.8 | Helped |
| Set latent-part count to 16 rather than smaller or larger tested counts (ablation) | FAPIS · COCO-20^0, K=1 | instance segmentation mAP50: 16.3, 17.2, 17.9, 18.4, 18.8, 18.5, 18.0 | Mixed |
| Use latent parts and their layouts rather than YOLACT prototypes | few-shot instance segmentation · COCO-20^i, K=1 and K=5 | mAP50: FAPIS improves by 2% over YOLACT | Helped |
Closest papers by meaning
- iFS-RCNN: An Incremental Few-Shot Instance Segmenter (CVPR 2022 · similarity 0.90)
- Incremental Few-Shot Instance Segmentation (CVPR 2021 · similarity 0.89)
- The Devil is in the Object Boundary: Towards Annotation-free Instance Segmentation using Foundation Models (ICLR 2024 · similarity 0.89)
- PartDistillation: Learning Parts From Instance Segmentation (CVPR 2023 · similarity 0.88)
- Beyond Pixel and Object: Part Feature as Reference for Few-Shot Video Object Segmentation (AAAI 2025 · similarity 0.88)
- Towards Open-World Segmentation of Parts (CVPR 2023 · similarity 0.88)
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.