AI Paper Atlas

Research map / Video / Action recognition

Motion Guided Attention Fusion To Recognize Interactions From Videos

Tae Soo Kim, Jonathan Jones, Gregory D. Hager

ICCV 2021 · Video › Action recognition · approach: Disentangled action recognition · 16 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
fuse object features with MGAF instead of concatenation (ablation)I3D single pathway · Something-Else compositionaltop-1 accuracy: 54.7 -> 60.5Improved
add object features via MGAF to single-pathway RGB modelI3D · Something-Else compositionaltop-1 accuracy: 46.8 -> 60.5Improved
fuse objects into the motion pathway with MGAF while keeping appearance pathwaySlowFast dual pathway · Something-Else compositionaltop-1 accuracy: 49.6 -> 68.0Improved
fuse MGAF objects into motion pathway instead of appearance pathway (ablation)dual pathway + MGAF · Something-Else compositionaltop-1 accuracy: 63.8 -> 68.0Improved
add motion pathway to appearance pathway (RGB only) (ablation)SlowFast · Something-Else compositionaltop-1 accuracy: 46.8 -> 49.6Improved
replace concatenation with MGAF for dual pathway plus objects (ablation)dual pathway · Something-Else compositionaltop-1 accuracy: 58.8 -> 68.0Improved
concatenate object features with appearance pathway (ablation)I3D · Something-Else compositionaltop-1 accuracy: 52.3 -> 54.7Improved
fuse MGAF objects into motion pathway only without appearance pathway (ablation)motion pathway · Something-Else compositionaltop-1 accuracy: 60.5 -> 55.8Hurt
use larger object-only model v2 over v1 (ablation)object temporal model · Something-Else compositionaltop-1 accuracy: 52.3 -> 53.6Improved
train and test on disjoint verb-noun compositions rather than mixedSlowFast · Something-Else compositionaltop-1 accuracy: 64.9 -> 49.6Hurt
use predicted detections instead of ground truth boxesMGAF(DP, Obj) · Something-Else compositionaltop-1 accuracy: 68.0 -> 61.2Hurt
add MGAF object fusion to dual pathwaySlowFast · IKEA-ASM original taskmicro accuracy: 73.5 -> 78.8Improved

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