AI Paper Atlas

Research map / Detection and segmentation / Weakly supervised detection and segmentation

Gradient-based Sampling for Class Imbalanced Semi-supervised Object Detection

Jiaming Li, Xiangru Lin, Wei Zhang, Xiao Tan, Yingying Li, Junyu Han, Errui Ding, Jingdong Wang, Guanbin Li

ICCV 2023 · Detection and segmentation › Weakly supervised detection and segmentation · approach: Pseudo-labeling for semi-supervised detection · 14 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 gradient-based reweighting (GbR) to Soft teacher (ablation)Soft teacher · MS-COCO sub-task split1AP (minority): 1.5 -> 7.9Improved
add class-rebalancing sampling (CrS) (ablation)Soft teacher · MS-COCO sub-task split1AP (minority): 1.5 -> 6.6Improved
add gradient-based thresholding (GbT) (ablation)Soft teacher · MS-COCO sub-task split1AP (minority): 1.5 -> 5.7Improved
combine CrS, GbT and GbR (ablation)Soft teacher · MS-COCO sub-task split1AP (all): 23.6 -> 26.5Improved
replace classification loss with focal loss (ablation)Soft teacher · MS-COCO sub-task split1AP (all): 23.6 -> 23.8No clear effect
add all modules versus Soft teacher in majority classes (ablation)Soft teacher · MS-COCO sub-task split1AP (majority): 31.0 -> 30.7No clear effect
use gradient-based thresholding instead of fixed 0.5 threshold (ablation)full method · MS-COCO sub-task split1AP (all): 25.6 -> 26.5Improved
combine both theta^p and score-based theta^c thresholds (ablation)GbT · MS-COCO sub-task split1AP (all): 26.4 -> 26.5Improved
use GbT instead of LabelMatch thresholding (ablation)full method · MS-COCO sub-task split1AP (all): 25.9 -> 26.5Improved
apply CReST class-imbalanced SSL classification method to Soft teacherSoft teacher · MS-COCO sub-task split1AP (all): 23.6 -> 22.2Hurt
apply the method on LVIS where unlabeled minority instances are scarceSoft teacher · LVIS sub-taskAP (all): 7.6 -> 8.9Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 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