AI Paper Atlas

Research map / Label-efficient and robust learning / Knowledge distillation

Few-Shot Object Detection by Knowledge Distillation Using Bag-of-Visual-Words Representations

Wenjie Pei, Shuang Wu, Dianwen Mei, Fanglin Chen, Jiandong Tian, Guangming Lu

ECCV 2022 · Label-efficient and robust learning › Knowledge distillation · approach: Knowledge distillation for detectors · 21 citations (OpenAlex)

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 changedWhereMeasuredEffect
add BoVW knowledge distillationTFA++ · PASCAL VOC Novel Split 1 (2-shot)nAP50: 42.1 -> 50.2Improved
add BoVW knowledge distillationTFA++ · PASCAL VOC Novel Split 1 (10-shot)nAP50: 60.8 -> 64.2Improved
add BoVW knowledge distillationDeFRCN · PASCAL VOC Novel Split 1 (2-shot)nAP50: 58.6 -> 62.5Improved
add BoVW knowledge distillationDeFRCN · PASCAL VOC Novel Split 3 (10-shot)nAP50: 62.5 -> 62.2No clear effect
add BoVW knowledge distillationDeFRCN · MS COCO (30-shot)nAP: 22.5 -> 22.6No clear effect
add BoVW knowledge distillationTFA++ · MS COCO (10-shot)nAP: 11.7 -> 12.8Improved
train with only 10% of base data, add BoVW distillationDeFRCN · MS COCO 10% base data (30-shot)nAP: 14.9 -> 16.0Improved
add base-class distillation (ablation)TFA++ · PASCAL VOC Novel Split 1 (3-shot)nAP50: 47.3 -> 51.5Improved
add novel-class distillation (ablation)TFA++ · PASCAL VOC Novel Split 1 (5-shot)nAP50: 59.7 -> 60.6Improved
add score fusion of original and BoVW classification (ablation)TFA++ · PASCAL VOC Novel Split 1 (5-shot)nAP50: 60.6 -> 62.1Improved
add DeCov loss (ablation)TFA++ · PASCAL VOC Novel Split 1 (10-shot)nAP50: 61.3 -> 63.1Improved
distill on BoVW representations instead of deep features (ablation)TFA++ · PASCAL VOC Novel Split 1 (10-shot)nAP50: 48.3 -> 64.2Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy