AI Paper Atlas

Research map / Graph learning / Knowledge and temporal graphs

Topic Scene Graph Generation by Attention Distillation From Caption

Wenbin Wang, Ruiping Wang, Xilin Chen

ICCV 2021 · Graph learning › Knowledge and temporal graphs · approach: Scene graph generation · 20 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 attention distillation from captions (weak supervision) to Up-Down captionerUD-ICRC · important relationship discoverymean recall: 12.64 -> 34.06Improved
add attention distillation from captions to Transformer captionerTransformer-ICRC · important relationship discoverymean recall: 9.82 -> 37.27Improved
use max pooling instead of mean pooling over time steps (ablation)UD-ICRC-attn · important relationship discoverymean recall: 12.84 -> 26.63Improved
mask non-noun words when collecting attention (ablation)UD-ICRC-attn · important relationship discoverymean recall: 34.06 -> 26.63Hurt
use subject, object and union features instead of union only (ablation)UD-ICRC-attn · important relationship discoverymean recall: 21.95 -> 33.76Improved
use subject, object and union features instead of subject and object only (ablation)UD-ICRC-attn · important relationship discoverymean recall: 25.98 -> 33.76Improved
add semantic embeddings of subject and object categories (ablation)UD-ICRC-attn · important relationship discoverymean recall: 33.76 -> 34.06Improved
add joint training with relational captions (lambda 0.1) to Up-Down captioner (ablation)UD · image captioningCIDEr-D: 94.1 -> 95.3Improved
add joint training with relational captions to Transformer captioner (ablation)Transformer · image captioningCIDEr-D: 85.6 -> 91.5Improved
increase relational caption loss weight lambda from 0.1 to 1.0 (ablation)UD-ICRC · image captioningCIDEr-D: 95.3 -> 93.5Hurt
increase relational caption loss weight lambda from 0.1 to 1.0 (ablation)UD-ICRC · important relationship discoveryR@20: 3.45 -> 2.31Hurt
add joint training with image captions to Transformer relational captioner (ablation)Transformer-RC · relational captioningmAP: 5.26 -> 5.15No clear effect

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