AI Paper Atlas

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

Logit Normalization for Long-Tail Object Detection

Liang Zhao, Yao Teng, Limin Wang

IJCV 2024 · Detection and segmentation › Open-set and few-shot detection · approach: Open-set and unknown object detection · 19 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 logit mean subtraction (ablation)LogN on Mask R-CNN R101 RFS · LVIS v1 valAPb: 27.5 -> 28.7Improved
add logit std division (ablation)LogN on Mask R-CNN R101 RFS · LVIS v1 valAPb: 27.5 -> 29.1Improved
add background calibration scalar beta (ablation)LogN on Mask R-CNN R101 RFS · LVIS v1 valAPb: 27.5 -> 29.0Improved
combine mean, std and beta (ablation)LogN on Mask R-CNN R101 RFS · LVIS v1 valAPb: 27.5 -> 29.8Improved
set beta to minimum foreground mean instead of background logit mean (ablation)LogN on Mask R-CNN R101 RFS · LVIS v1 valAPb: 12.1 -> 29.8Improved
set beta to minimum foreground mean instead of maximum foreground mean (ablation)LogN on Mask R-CNN R101 RFS · LVIS v1 valAPb: 25.5 -> 29.8Improved
set beta to minimum instead of average foreground mean (ablation)LogN on Mask R-CNN R101 RFS · LVIS v1 valAPb: 29.4 -> 29.8Improved
adaptive beta instead of constant -12 (ablation)LogN on Mask R-CNN R101 RFS · LVIS v1 valAPb: 29.3 -> 29.8Improved
online LogN during training with beta=0 instead of post-hoc (ablation)LogN on LVIS detector · LVIS v1 valper-group APHurt
square the standard deviation in LogN (ablation)LogN on Mask R-CNN R101 RFS · LVIS v1 valAP: 23.1% rare AP vs 23.2; total -0.5%Mixed
replace LogA with LogN under RFSCascade Mask R-CNN R101 · LVIS v1 valAPb: 32.7 -> 33.8Improved
replace Seesaw Loss with LogN under RFSCascade Mask R-CNN R101 · LVIS v1 valAPb: 32.8 -> 33.8Improved

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