AI Paper Atlas

Research map / Detection and segmentation / Open-set and few-shot detection

SKDF: A Simple Knowledge Distillation Framework for Distilling Open-Vocabulary Knowledge to Open-World Object Detector

Shuailei Ma, Yuefeng Wang, Ying Jin Wei, Enming Zhang, Jiaqi Fan, Xinyu Sun, Peihao Chen

TPAMI 2025 · Detection and segmentation › Open-set and few-shot detection · approach: Open-vocabulary object detection · 18 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
distill open-world knowledge from GLIP instead of using GLIP directlyOWOD detector · OWOD split Task 1U-Recall: 37.0 -> 39.0Improved
distill from GLIP instead of using GLIP directlyOWOD detector · MS-COCO split Task 1U-Recall: 52.6 -> 60.9Improved
add naive knowledge distillation without down-weight or cascade structure (ablation)GLIP baseline · OWOD split Task 2known mAP (both): 32.4 -> 19.2Hurt
add down-weight loss to distillation (ablation)distilled detector · OWOD split Task 1known mAP (current): 41.1 -> 53.5Improved
add cascade decoupled structure to distillation (ablation)distilled detector · OWOD split Task 1known mAP (current): 41.1 -> 51.9Improved
combine down-weight loss and cascade structure (ablation)distilled detector · OWOD split Task 2known mAP (both): 34.9 -> 40.3Improved
add down-weight and cascade components (ablation)distilled detector · OWOD split Task 1unknown recall: 39.4 -> 39.0No clear effect
use two separate decoders instead of the standard DDETR decoder (ablation)SKDF · OWOD splitknown mAP: 53.5 -> 56.8Improved
use two separate decoders instead of CAT shared decoder (ablation)SKDF · OWOD splitknown mAP: 55.3 -> 56.8Improved
distill from GLIP instead of SAM (ablation)SKDF · OWOD splitknown mAP: 19.4 -> 56.8Improved
use LVIS text prompt instead of COCO prompt for distillation (ablation)SKDF · OWOD splitunknown recall (UR sam): 12.9 -> 16.5Improved
add UnSniffer unknown post-processing to SKDFSKDF · IntensiveSetunknown precision: 16.2 -> 37.9Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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