AI Paper Atlas

Research map / Video / Action recognition

Action Genome: Actions As Compositions of Spatio-Temporal Scene Graphs

CVPR 2020 · Video › Action recognition · approach: Disentangled action recognition · 309 citations (OpenAlex)

Check your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's title only (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 spatio-temporal scene graph feature banks to LFB-style action recognitionAction recognition on Charades · Charades validationmAP: +1.8% mAPHelped
Use ground-truth scene graphs instead of predicted scene graphs (ablation)SGFB · Charades validationmAP: 16% increase on mAPHelped
Add spatio-temporal scene graph featuresFew-shot action recognition · Charades novel classes, 1-shot, 5-shot, and 10-shotmAP: 28.3 vs 28.8; 36.3 vs 37.9; 39.6 vs 42.7Helped
Use ground-truth scene graphs instead of predicted scene graphs (ablation)SGFB few-shot recognition · Charades novel classes, 10-shotmAP: 10.9% 10-shot mAP improvementHelped
Use video frames rather than single images as inputRelDN scene graph prediction · Action Genome, PredCls, SGCls, and SGGenRecall at 20 and 50: PredCls R@20: 49.37 -> 48.80; R@50: 49.58 -> 48.98; SGCls R@20: 46.74 -> 46.19; R@50: 49.36 -> 48.76; SGGen R@20: 35.61 Hurt

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