Research map / Detection and segmentation / Open-set and few-shot detection
Hallucination Improves 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 |
|---|---|---|---|
| Add feature-space hallucination | CoRPNs w/ cos · COCO novel classes, 1–3 shot | AP: 4.1 -> 4.4; 5.4 -> 5.6; 7.1 -> 7.2 | Helped |
| Add feature-space hallucination | TFA w/ cos · COCO novel classes, 1–3 shot | AP: 3.4 -> 3.8; 4.6 -> 5.0; 6.6 -> 6.9 | Helped |
| Add feature-space hallucination | CoRPNs w/ cos · PASCAL VOC novel sets 1–3, 1-shot | AP50: 44.4 -> 47.0; 25.7 -> 26.3; 35.8 -> 40.4 | Helped |
| Add feature-space hallucination | TFA w/ cos · PASCAL VOC novel sets 1–3, 1-shot | AP50: 39.8 -> 45.1; 23.5 -> 23.2; 30.8 -> 30.5 | Mixed |
| Use EM-style alternating training instead of joint training (ablation) | CoRPNs with hallucination · PASCAL VOC novel sets 1–3, 1-shot | AP50: 38.5 -> 46.7; 22.2 -> 24.5; 33.6 -> 38.5 | Helped |
| Add a second EM iteration (ablation) | CoRPNs with hallucination · PASCAL VOC novel sets 1–3, 1-shot | AP50: 46.7 -> 47.0; 24.5 -> 26.3; 38.5 -> 40.4 | Helped |
| Use more than two EM iterations (ablation) | CoRPNs with hallucination · Few-shot detection | AP50 | No clear effect |
| Use the aggressive hallucinator rather than the conservative variant (ablation) | TFA · PASCAL VOC novel sets 1–3, 1-shot | AP50: 45.1 -> 40.8; 23.2 -> 29.3; 30.5 -> 33.7 | Mixed |
| Vary hallucinated examples per class from 0 to 20 (ablation) | CoRPNs + Halluc · PASCAL VOC novel set 1, 1-shot | AP50: 44.4; 47.1; 47.0; 48.0; 47.9; 47.8; 47.0 | Mixed |
| Add hallucination with a fully-connected classifier (ablation) | CoRPNs w/ fc · PASCAL VOC novel sets 1–3, 1-shot | AP50: 40.8 -> 44.2; 20.4 -> 23.0; 29.4 -> 31.5 | Helped |
| Add feature-space hallucination | CoRPNs w/ cos · COCO base classes after fine-tuning, 1–3 shot | AP: 34.1 -> 32.3; 34.7 -> 34.5; 34.8 -> 34.7 | Hurt |
| Add feature-space hallucination | TFA w/ cos · COCO base classes after fine-tuning, 1–3 shot | AP: 34.1 -> 31.5; 34.7 -> 32.8; 34.7 -> 33.3 | Hurt |
Closest papers by meaning
- Low-Shot Learning From Imaginary Data (CVPR 2018 · similarity 0.88)
- Label Hallucination for Few-Shot Classification (AAAI 2022 · similarity 0.88)
- Few-Shot Object Detection With Attention-RPN and Multi-Relation Detector (CVPR 2020 · similarity 0.88)
- GLSim: Detecting Object Hallucinations in LVLMs via Global-Local Similarity (NeurIPS 2025 · similarity 0.86)
- Few-Shot Learner Generalizes Across AI-Generated Image Detection (ICML 2025 · similarity 0.85)
- Beyond the Global Scores: Fine-Grained Token Grounding as a Robust Detector of LVLM Hallucinations (CVPR 2026 · 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.