AI Paper Atlas

Research map / Label-efficient and robust learning / Self-supervised and contrastive

On Compositions of Transformations in Contrastive Self-Supervised Learning

Mandela Patrick, Yuki M. Asano, Polina Kuznetsova, Ruth Fong, João F. Henriques, Geoffrey Zweig, Andrea Vedaldi

ICCV 2021 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Inductive bias and minimal representations · 148 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 audio modality (cross-modal instead of V-only SimCLR-like) (ablation)GDT · HMDB-51finetune accuracy: 44.6 -> 52.4Improved
add time-shift and time-reversal invariance in visual-only setting (ablation)SimCLR-like video · HMDB-51finetune accuracy: 44.6 -> 37.8Hurt
add TS and TR invariance in cross-modal setting (ablation)GDT cross-modal · HMDB-51finetune accuracy: 52.4 -> 59.9Improved
switch TS to distinctiveness (keep TR invariant) (ablation)GDT cross-modal · HMDB-51finetune accuracy: 59.9 -> 61.4Improved
make both TS and TR distinctive (ablation)GDT cross-modal · HMDB-51finetune accuracy: 61.4 -> 57.2Hurt
make both TS and TR distinctive (ablation)GDT cross-modal · HMDB-51frozen retrieval R@1: 27.1 -> 20.5Hurt
learn invariance rather than ignoring TS (ablation)GDT cross-modal · HMDB-51finetune accuracy: 57.4 -> 59.9Improved
cross-modal instead of within-modal contrast (ablation)GDT · HMDB-51finetune accuracy: 37.8 -> 52.4Improved
cross-modal instead of within-modal contrast (ablation)GDT · UCF-101finetune accuracy: 76.4 -> 87.6Improved
video-text cross-modal instead of SimCLR visual only (ablation)GDT · HMDB-51 (HowTo100M)accuracy: 36.1 -> 59.2Improved
add TS distinctiveness and TR invariance to video-text (ablation)GDT video-text · HMDB-51 (HowTo100M)accuracy: 59.2 -> 64.4Improved
add TS+TR invariance in visual-only setting (ablation)SimCLR-like video · HMDB-51frozen retrieval R@1: 11.8 -> 13.9Improved

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