AI Paper Atlas

Research map / Video / Action recognition

How Transferable are Video Representations Based on Synthetic Data?

Yo-whan Kim, Samarth Mishra, SouYoung Jin, Rameswar Panda, Hilde Kuehne, Leonid Karlinsky, Venkatesh Saligrama, Kate Saenko, Aude Oliva, Rogerio Feris

NeurIPS 2022 · Video › Action recognition · approach: Disentangled action recognition · 16 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
pre-train on synthetic videos instead of KineticsTSN ResNet-50 · UCF101 and HMDB51 (high bias) fine-tuningtop-1 accuracy: 86.17 -> 83.40 (UCF101), 57.45 -> 54.38 (HMDB51)Hurt
pre-train on synthetic videos instead of KineticsTSN ResNet-50 · Mini-SSV2, Diving48, IkeaFA, UAV-Human (low bias) fine-tuningtop-1 accuracy: 48.50 -> 49.69 (Mini-SSV2), 32.45 -> 35.57 (UAV-Human)Improved
pre-train on synthetic videos instead of KineticsTSN ResNet-50 · UCF101 linear probingtop-1 accuracy: 69.92 -> 28.02Hurt
pre-train on synthetic videos instead of KineticsTimeSformer · UAV-Human fine-tuningtop-1 accuracy: 23.27 -> 24.95Improved
pre-train on synthetic videos instead of KineticsTimeSformer · Diving48 fine-tuningtop-1 accuracy: 46.37 -> 44.86Hurt
pre-train on synthetic videos instead of KineticsR(2+1)D ResNet-50 · UCF101 fine-tuningtop-1 accuracy: 87.21 -> 80.02Hurt
use 16 input frames instead of 8 for synthetic pre-training (ablation)TSN synthetic · UAV-Human fine-tuningtop-1 accuracy: 35.57 -> 40.51Improved
increase number of pre-training classes (30, 90, 150) (ablation)Kinetics and synthetic pre-training · UCF101, HMDB51, UAV-Humantop-1 accuracyImproved
increase number of pre-training classes (30, 90, 150) (ablation)Kinetics and synthetic pre-training · Mini-SSV2 and Diving48top-1 accuracyNo clear effect
increase samples per class (250, 750, 1000) (ablation)Kinetics and synthetic pre-training · six downstream taskstop-1 accuracyNo clear effect
replace Kinetics with Moments-in-Time real pre-training (ablation)TimeSformer · UCF101 fine-tuningtop-1 accuracy: 92.09 -> 91.24No clear effect

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