AI Paper Atlas

Research map / Video / Action recognition

Grouped Spatial-Temporal Aggregation for Efficient Action Recognition

Chenxu Luo, Alan L. Yuille

ICCV 2019 · Video › Action recognition · approach: Spatiotemporal CNNs for action recognition · 175 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
replace one group of 3D group convolution with spatial-only convolution (GST-Large vs C3D groups=2)C3D groups=2 · Something-Something v1 valtop-1 accuracy: 45.1 -> 47.7Improved
parallel spatial/temporal channel split instead of cascaded P3D decouplingP3D · Something-Something v1 valtop-1 accuracy: 45.7 -> 47.7Improved
same parallel split on v2P3D · Something-Something v2 valtop-1 accuracy: 59.8 -> 62.0Improved
GST-Large versus full 3x3x3 C3D with fewer parametersC3D 3x3x3 · Something-Something v1 valtop-1 accuracy: 46.2 -> 47.7Improved
add GST temporal channels to 2D ResNet-50 at equal parametersC2D · Something-Something v1 valtop-1 accuracy: 20.4 -> 46.7Improved
reduce temporal channel ratio alpha from 1/2 to 1/4 (ablation)GST · Something-Something v1 valtop-1 accuracy: 46.7 -> 47.0Improved
reduce temporal channel ratio alpha from 1/4 to 1/8 (ablation)GST · Something-Something v2 valtop-1 accuracy: 61.6 -> 60.7Hurt
reduce temporal channel ratio alpha from 1/2 to 1/8 (ablation)GST · Something-Something v1 valtop-1 accuracy: 46.7 -> 46.7No clear effect
increase input frames from 8 to 16GST · Something-Something v1 valtop-1 accuracy: 47.0 -> 48.6Improved
increase input frames from 8 to 16GST · Something-Something v2 valtop-1 accuracy: 61.6 -> 62.6Improved
use two clips at test time instead of oneGST · Something-Something v1 valtop-1 accuracy: 47.0 -> 47.6Improved
GST instead of P3D backboneResNet-50 · Diving48accuracy: 32.4 -> 38.8Improved

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