AI Paper Atlas

Research map / Label-efficient and robust learning / OOD and anomaly detection

Revisiting Out-of-Distribution Detection in Real-Time Object Detection: From Benchmark Pitfalls to a New Mitigation Paradigm

Changshun Wu, Weicheng He, Chih-Hong Cheng, Xiaowei Huang, Saddek Bensalem

TPAMI 2026 · Label-efficient and robust learning › OOD and anomaly detection · approach: Spurious-aware OOD detection · 3 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
fine-tune on proximal OoD samples treated as backgroundYOLOv10 · BDD-100K Near-OoDOoD detection count: 701 -> 133Improved
add KNN OoD filter to the fine-tuned detectorfine-tuned YOLOv10 · BDD-100K Near-OoDOoD detection count: 133 -> 80Improved
fine-tune on proximal OoD samplesFaster R-CNN · BDD-100K Near-OoDOoD detection count: 2576 -> 394Improved
fine-tune on proximal OoD samplesRT-DETR · BDD-100K Near-OoDOoD detection count: 3145 -> 967Improved
add KNN filter to fine-tuned RT-DETRfine-tuned RT-DETR · PASCAL-VOC Far-OoDOoD detection count: 966 -> 470Improved
discard proposals whose background score is highest at inference (ablation)Faster R-CNN · PASCAL-VOCmAP: 0.402 -> 0.391Mixed
remove ID outlier objects from calibration/test ID data (ablation)KNN filter · PASCAL-VOC vs OpenImagesFPR95: 0.5376 -> 0.4467Improved
remove ID-labelled objects from the OoD test set (fix Type 1 error) (ablation)OoD benchmark evaluation · PASCAL-VOC vs OpenImagesFPR95: +9.85% inflation removedImproved
use error-prone ID subset (2,000 images) instead of all ID data (ablation)proximal OoD fine-tuning · BDD-100KmAP / Near-OoD count: 0.2556 -> 0.2492; 133 -> 170Mixed
use Object365 instead of Caltech256 as proximal OoD source (ablation)proximal OoD fine-tuning (ID_err) · BDD-100KNear-OoD count: 170 -> 118Improved
select proximal categories by closest CLIP text similarity instead of farthest (ablation)proximal OoD fine-tuning · BDD-100KNear-OoD count: 226 -> 143Improved
closer proximal OoD data (ablation)proximal OoD fine-tuning · BDD-100KmAP (all): 0.2523 -> 0.2457Hurt

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