AI Paper Atlas

Research map / Video / Action recognition

Something-Else: Compositional Action Recognition With Spatial-Temporal Interaction Networks

Joanna Materzynska, Tete Xiao, Roei Herzig, Huijuan Xu, Xiaolong Wang, Trevor Darrell

CVPR 2020 · Video › Action recognition · approach: Disentangled action recognition · 164 citations (OpenAlex)

Check 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 object identity embeddings to STIN (ablation)STIN · Something-Else compositional split with ground-truth boxestop-1 accuracy: 47.1 -> 51.3Improved
Add object identity embeddings to STIN (ablation)STIN · Something-Else compositional split with detected boxestop-1 accuracy: 34.1 -> 36.7Improved
Add non-local aggregation to STIN with OIE (ablation)STIN + OIE · Something-Else compositional split with ground-truth boxestop-1 accuracy: 51.3 -> 51.4No clear effect
Add non-local aggregation to STIN with OIE (ablation)STIN + OIE · Something-Else compositional split with detected boxestop-1 accuracy: 36.7 -> 37.2Improved
Add object identity embeddings to STIN (ablation)STIN · Something-Else few-shot novel categories, ground-truth boxes, 5-shotaccuracy: 24.5 -> 25.8Improved
Add non-local aggregation to STIN with OIE (ablation)STIN + OIE · Something-Else few-shot novel categories, ground-truth boxes, 5-shotaccuracy: 25.8 -> 27.7Improved
Add object identity embeddings to STIN (ablation)STIN · Something-Else few-shot novel categories, detected boxes, 5-shotaccuracy: 14.2 -> 16.3Improved
Add non-local aggregation to STIN with OIE (ablation)STIN + OIE · Something-Else few-shot novel categories, detected boxes, 5-shotaccuracy: 16.3 -> 17.7Improved
Train on only one object category, box (ablation)STIN + OIE · Something-Else one-class compositional split with ground-truth boxestop-1 accuracy: 14.7 -> 28.5Improved
Combine I3D appearance features with STIN interaction featuresI3D · Something-Something V2 original splittop-1 accuracy: 55.5 -> 60.2Improved
Use an ensemble of I3D and STIN rather than I3D aloneI3D · Something-Else compositional split with ground-truth boxestop-1 accuracy: 46.8 -> 58.1Improved
Use an ensemble of STRG and STINSTRG · Something-Else compositional split with detected boxestop-1 accuracy: 52.3 -> 56.2Improved

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