AI Paper Atlas

Research map / Video / Video segmentation and tracking

Self-Supervised Video Object Segmentation by Motion Grouping

Charig Yang, Hala Lamdouar, Erika Lu, Andrew Zisserman, Weidi Xie

ICCV 2021 · Video › Video segmentation and tracking · approach: Matching-based video object segmentation · 153 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 PWC-Net flow with the proposed model instead of CIS (ablation)motion grouping model · DAVIS2016J mean: outperforms CIS by about 4.5 pointsImproved
use RAFT flow instead of PWC-Net (ablation)motion grouping model · DAVIS2016J meanImproved
add instance normalization (ablation)motion grouping model · DAVIS2016J meanImproved
increase grouping iterations from T=3 to T=5 (ablation)motion grouping model · DAVIS2016J meanImproved
increase grouping iterations to T=8 (ablation)motion grouping model · DAVIS2016J meanHurt
add temporal consistency loss (ablation)motion grouping model · DAVIS2016J meanImproved
add entropy regularization (ablation)motion grouping model · DAVIS2016J meanImproved
self-supervised flow-only grouping vs CIS (no post-processing)motion grouping model · DAVIS2016J mean: +9.1%Improved
use flow-only grouping vs CIS on camouflagemotion grouping model · MoCAsuccess rate: 14%Improved
use flow-only grouping vs supervised CODmotion grouping model · MoCAJ: 18.5% on JaccardImproved

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