AI Paper Atlas

Research map / Video / Action recognition

Video-FocalNets: Spatio-Temporal Focal Modulation for Video Action Recognition

Syed Talal Wasim, Muhammad Uzair Khattak, Muzammal Naseer, Salman Khan, Mubarak Shah, Fahad Shahbaz Khan

ICCV 2023 · Video › Action recognition · approach: Spatiotemporal CNNs for action recognition · 47 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
use Video-FocalNet instead of Video-Swin at Tiny sizeVideo-FocalNet-T · Kinetics-400 valtop-1 accuracy: 78.8 -> 79.8Improved
use Video-FocalNet instead of Video-Swin at Small sizeVideo-FocalNet-S · Kinetics-400 valtop-1 accuracy: 80.6 -> 81.4Improved
use Video-FocalNet-B with ImageNet-1K init instead of Uniformer-BVideo-FocalNet-B · Kinetics-400 valtop-1 accuracy: 83.0 -> 83.6Improved
use Video-FocalNet-B instead of MViTv2-BVideo-FocalNet-B · Kinetics-600 valtop-1 accuracy: 85.5 -> 86.7Improved
use Video-FocalNet-B instead of TFCNETVideo-FocalNet-B · Diving-48 V2top-1 accuracy: 88.3 -> 90.8Improved
use Video-FocalNet-B instead of Video-Swin-BVideo-FocalNet-B · ActivityNet 1.3top-1 accuracy: 88.5 -> 89.8Improved
fuse spatial and temporal modulators by elementwise multiplication instead of averaging or learnable projection (ablation)Video-FocalNet-S · Kinetics-400 valtop-1 accuracyImproved
use simple patch embedding (T=1) instead of tubelet embedding (T=2) (ablation)Video-FocalNet-S · Kinetics-400 valtop-1 accuracyImproved
use parallel spatial and temporal focal modulation instead of factorized, divided or averaged designs (ablation)Video-FocalNet-S · Kinetics-400 valtop-1 accuracy and FLOPsImproved

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