AI Paper Atlas

Research map / Video / Action recognition

In Defense of Image Pre-training for Spatiotemporal Recognition

Xianhang Li, Huiyu Wang, Chen Wei, Jieru Mei, Alan Yuille, Yuyin Zhou, Cihang Xie

ECCV 2022 · Video › Action recognition · approach: Spatiotemporal CNNs for action recognition · 2 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
add ImageNet image pre-training (center-kernel init) (ablation)ResNet50-3x3x3 on Kinetics-400 · Kinetics-400 (fixed budget)top-1 accuracy: 70.7 -> 74.3Improved
add ImageNet image pre-training (ablation)ResNet50-3x1x1 on Kinetics-400 · Kinetics-400 (fixed budget)top-1 accuracy: 73.1 -> 74.9Improved
add STS convolution on top of image pre-training (ablation)ResNet50-3x3x3 · Kinetics-400 (fixed budget)top-1 accuracy: 74.3 -> 74.7Improved
add image pre-training and STS conv to SS-V2 modelsSlowfast50 (4x16)-32f · Something-Something V2top-1 accuracy: 49.9 -> 57.2Improved
add STS conv on top of image pre-training (ablation)Five 3D CNNs · Something-Something V2top-1 accuracy: +1.2% averageImproved
add image pre-training and STS conv to R2plus1D34-8fR2plus1D34-8f · Kinetics-400top-1 accuracy: 68.7 -> 76.2Improved
add image pre-training and STS conv to X3D-S-13fX3D-S-13f · Kinetics-400top-1 accuracy: 73.2 -> 73.5No clear effect
add ImageNet pre-training at 32 frames (ablation)ResNet50-STS 3x3x3 · Kinetics-400 (32 frames)top-1 accuracy: 74.4 -> 76.7Improved
pre-train on K400 frames instead of ImageNet (ablation)ResNet50-STS 3x3x3 · Kinetics-400 (8 frames)top-1 accuracy: 74.7 -> 73.4Hurt
pre-train on K400 frames instead of ImageNet (ablation)X3D-S · Kinetics-400 (13 frames)top-1 accuracy: 73.5 -> 72.8Hurt
use zero-init instead of inflation (ablation)ResNet50-3x3x3 · Kinetics-400top-1 accuracy: 74.0 -> 74.3Improved
use zero-init instead of inflation (ablation)ResNet50-3x1x1 · Something-Something V2top-1 accuracy: 61.8 -> 61.6Mixed

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