Research map / Detection and segmentation / Open-set and few-shot detection
Incremental Few-Shot Object Detection
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 ONCE instead of Feature-Reweight for incremental few-shot detection | CentreNet/ResNet50 · COCO val2017, 1-shot, novel classes | AP: 0.1 -> 0.7 | Helped |
| Use ONCE instead of Feature-Reweight for incremental few-shot detection | CentreNet/ResNet50 · COCO val2017, 5-shot, novel classes | AP: 0.8 -> 1.0 | Helped |
| Use ONCE instead of Feature-Reweight for incremental few-shot detection | CentreNet/ResNet50 · COCO val2017, 10-shot, novel classes | AP: 1.5 -> 1.2 | Hurt |
| Use ONCE rather than Fine-Tuning to register novel classes | CentreNet/ResNet50 · DeepFashion2 novel classes, 1-shot incremental batch | AP: 2.8 -> 4.6 | Helped |
| Use ONCE rather than Fine-Tuning to register novel classes | CentreNet/ResNet50 · DeepFashion2 novel classes, 5-shot incremental batch | AP: 17.0 -> 29.5 | Helped |
| Use ONCE rather than Fine-Tuning to register novel classes | CentreNet/ResNet50 · DeepFashion2 novel classes, 10-shot incremental batch | AP: 17.1 -> 32.2 | Helped |
| Use class-specific codes and independent class detection to add novel classes | ONCE · COCO val2017, base classes across 1-, 5-, and 10-shot incremental batch | AP: 17.9; 17.9; 17.9 | No clear effect |
| Register novel classes one at a time instead of in a single batch | ONCE · COCO continuous incremental learning | all-classes accuracy | No clear effect |
| Use ONCE rather than MAML for cross-dataset few-shot detection | CentreNet/ResNet50 · VOC2007 test, 5-shot incremental batch, novel classes | AP: 0.6 -> 2.4 | Helped |
| Use ONCE rather than MAML for cross-dataset few-shot detection | CentreNet/ResNet50 · VOC2007 test, 10-shot incremental batch, novel classes | AP: 1.0 -> 2.6 | Helped |
| Use ONCE rather than Meta R-CNN | COCO val2017, non-incremental 10-shot novel-class detection · COCO val2017 | AP: 8.7 -> 5.1 | Hurt |
Closest papers by meaning
- Incremental-DETR: Incremental Few-Shot Object Detection via Self-Supervised Learning (AAAI 2023 · similarity 0.93)
- Sylph: A Hypernetwork Framework for Incremental Few-Shot Object Detection (CVPR 2022 · similarity 0.91)
- iFS-RCNN: An Incremental Few-Shot Instance Segmenter (CVPR 2022 · similarity 0.89)
- Revisiting Generative Replay for Class Incremental Object Detection (CVPR 2025 · similarity 0.88)
- Fine-Grained Prototypes Distillation for Few-Shot Object Detection (AAAI 2024 · similarity 0.87)
- Revisiting Few-Shot Object Detection with Vision-Language Models (NeurIPS 2024 · similarity 0.87)
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.