AI Paper Atlas

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

Self-Supervised Any-Point Tracking by Contrastive Random Walks

Ayush Shrivastava, Andrew Owens

ECCV 2024 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Invariance-based self-supervised learning · 3 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 label warping to the contrastive random walk with a transformer (ablation)GMRW · TAP-Vid KubricAJ: 25.4 -> 45.3Improved
add label warping (ablation)GMRW · TAP-Vid DAVISAJ: 10.4 -> 32.1Improved
add edge-aware smoothness loss (ablation)GMRW · TAP-Vid KubricAJ: 45.3 -> 49.0Improved
lower training stride from 4 to 2 (ablation)GMRW · TAP-Vid DAVISAJ: 33.0 -> 34.5Improved
train on Kinetics instead of Kubric (ablation)GMRW · TAP-Vid KubricAJ: Kubric 47.7 -> 47.5; DAVIS 34.5 -> 34.6Mixed
decrease evaluation stride from 4 to 1 (ablation)GMRW · TAP-Vid DAVISAJ: 23.5 -> 41.8Improved
supervise the same architecture with ground-truth tracks instead of self-supervision (ablation)GMRW architecture · TAP-Vid DAVISAJ: supervised 39.1 vs TAP-Net 38.4Improved
GMRW versus self-supervised baselines across benchmarksGMRW-C · TAP-Vid KineticsAJ: GMRW-C 31.9 vs Flow-Walk-C 40.9 on Kinetics; 41.8 vs 35.2 on DAVISMixed

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