AI Paper Atlas

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

PseCo: Pseudo Labeling and Consistency Training for Semi-Supervised Object Detection

Gang Li, Xiang Li, Yujie Wang, Yichao Wu, Ding Liang, Shanshan Zhang

ECCV 2022 · Detection and segmentation › Weakly supervised detection and segmentation · approach: Pseudo-labeling for semi-supervised detection · 169 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 label-level consistency via random-resize scale jitter (ablation)PseCo · COCO val2017 (10% labeled, single fold)mAP: 32.7 -> 33.9Improved
add feature-consistency view V2 on top of V1 (ablation)PseCo · COCO val2017 (10% labeled, single fold)mAP: 33.9 -> 34.9Improved
add extra downsampled view with only label consistency (vanilla multi-view) (ablation)PseCo · COCO val2017 (10% labeled, single fold)mAP: 33.9 vs 33.9No clear effect
add PCV regression weighting to V1 (ablation)PseCo · COCO val2017 (10% labeled, single fold)mAP: 33.9 -> 34.8Improved
add PLA after PCV (ablation)PseCo · COCO val2017 (10% labeled, single fold)mAP: 34.8 -> 35.7Improved
add PLA to V1+V2+PCV (ablation)PseCo · COCO val2017 (10% labeled, single fold)mAP: 36.0 -> 36.3Improved
use PCV instead of abandoning regression loss (ablation)PseCo · COCO val2017 (10% labeled, single fold)mAP: 33.9 -> 34.8Improved
use PCV instead of box jittering (ablation)PseCo · COCO val2017 (10% labeled, single fold)mAP: 34.5 -> 34.8Improved
use PCV instead of reg consistency (ablation)PseCo · COCO val2017 (10% labeled, single fold)mAP: 34.2 -> 34.8Improved
combine score and IoU in PLA quality (alpha 0.5 vs 0) (ablation)PseCo · COCO val2017 (10% labeled, single fold)mAP: 35.2 -> 35.7Improved
combine score and IoU in PLA quality (alpha 0.5 vs 1.0) (ablation)PseCo · COCO val2017 (10% labeled, single fold)mAP: 35.4 -> 35.7Improved
raise PLA candidate IoU threshold t from 0.4 to 0.5 (ablation)PseCo · COCO val2017 (10% labeled, single fold)mAP: 35.7 -> 35.5Hurt

Later papers that compared with this method

9 later accepted papers report Pseco as a baseline in their results tables; 7 of them report a better number on most of the results they share with it.

1 more not listed.

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