AI Paper Atlas

Research map / Video / Video-language understanding

Prune Spatio-temporal Tokens by Semantic-aware Temporal Accumulation

Shuangrui Ding, Peisen Zhao, Xiaopeng Zhang, Rui Qian, Hongkai Xiong, Qi Tian

ICCV 2023 · Video › Video-language understanding · approach: Long video understanding · 23 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
apply STA token pruning to off-the-shelf ViT-H without retrainingViT-H · Kinetics-400Top-1 accuracy / GFLOPs: GFLOPs 1192 -> 611; accuracy 86.3 -> 86.1Improved
apply STA to VideoSwin-B with the same backbone vs STTSVideoSwin-B · Kinetics-400Top-1 accuracy: 81.9 -> 82.3 (149 vs 253 GFLOPs)Improved
apply STA to VideoSwin-B vs STTSVideoSwin-B · Something-Something V2Top-1 accuracy: 68.7 -> 69.2 (188 vs 237 GFLOPs)Improved
increase first-stage drop number r1 for small models (ViT-S to 64) (ablation)ViT-S · Kinetics-400Top-1 accuracy: 78.8 -> 78.1Hurt
combine temporal accumulation score with semantic score instead of either alone (ablation)STA on ViT-L · Kinetics-400Top-1 accuracy: 84.69 / 84.87 -> 84.96Improved
use decreasing drop schedule instead of constant (ablation)STA on ViT-L · Kinetics-400Top-1 accuracy: 84.68 -> 84.96Improved
use increasing drop schedule instead of decreasing (ablation)STA on ViT-L · Kinetics-400Top-1 accuracy: 84.96 -> 77.68Hurt
alternate accumulation order (forward-backward-forward) instead of the same direction (ablation)STA on ViT-L · Kinetics-400Top-1 accuracy: 84.26 -> 84.96Improved
choose first-frame pruning method (random, grid, ToMe) (ablation)STA on ViT-L · Kinetics-400Top-1 accuracy: 84.78 -> 84.96No clear effect
use key projection f_K for token affinity instead of f_Q, f_V or f_FFN (ablation)STA on ViT-L · Kinetics-400Top-1 accuracy: 84.83 -> 84.96No clear effect
train ViT-B with STA to halve training timeViT-B · Kinetics-400Top-1 accuracy: 81.2 -> 80.6Hurt

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