AI Paper Atlas

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

Learning To Generate Language-Supervised and Open-Vocabulary Scene Graph Using Pre-Trained Visual-Semantic Space

Yong Zhang, Yingwei Pan, Ting Yao, Rui Huang, Tao Mei, Chang-Wen Chen

CVPR 2023 · Detection and segmentation › Open-set and few-shot detection · approach: Open-vocabulary 3D and world perception · 41 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
Use VS3 with GLIP-L (Swin-L) instead of the strongest prior closed-set modelFully supervised SGG · VG150 test, SGDETR@100: 38.1 -> 41.5Improved
Transfer a pre-trained visual-semantic space to language-supervised SGGVS3 · VG150 test, unlocalized graph supervisionR@100: 13.15 -> 28.19Improved
Transfer a pre-trained visual-semantic space to language-supervised SGGVS3 · VG150 test, VG caption supervisionR@100: 12.14 -> 20.54Improved
Transfer a pre-trained visual-semantic space to language-supervised SGGVS3 · VG150 test, COCO caption supervisionR@100: 7.62 -> 9.90Improved
Use a larger pre-trained backbone, Swin-L instead of Swin-TVS3 · VG150 test, fully supervised SGDETR@100: 39.2 -> 41.5Improved
Add visual region features to relation representations (ablation)VS3 · VG150 test, fully supervised SGDETR@100: 36.7 -> 39.2Improved
Add spatial features to relation representations (ablation)VS3 · VG150 test, fully supervised SGDETR@100: 37.8 -> 39.2Improved
Use frequency bias for predicate classification (ablation)VS3(Swin-T) · VG150 test, language-supervised SGDET with unlocalized graphsR@100: 28.19 -> 31.75Improved
Use all captions per image instead of one caption (ablation)VS3 with simple parser · VG150 test, language-supervised SGDET on COCO captionsR@20/50/100: 5.07 / 6.25 / 7.36 -> 5.42 / 6.82 / 7.93Improved
Use the advanced parser instead of the simple parser (ablation)VS3 trained on all COCO captions · VG150 test, language-supervised SGDETR@20/50/100: 5.42 / 6.82 / 7.93 -> 5.59 / 7.30 / 8.62Improved
Use a larger pre-trained backbone, Swin-L instead of Swin-TVS3 · VG150 Ov-SGG PREDCLSR@50/100: 50.10 / 52.05 -> 55.88 / 58.18Improved
Train with VG captions rather than COCO captionsVS3(Swin-L) · VG150 language-supervised Ov-SGGR@50/100: 6.76 / 8.45 -> 12.98 / 16.29Improved

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