AI Paper Atlas

Research map / Label-efficient and robust learning / Semi-supervised learning

Multiview Pseudo-Labeling for Semi-Supervised Learning From Video

Bo Xiong, Haoqi Fan, Kristen Grauman, Christoph Feichtenhofer

ICCV 2021 · Label-efficient and robust learning › Semi-supervised learning · approach: Pseudo-labeling for specialized tasks · 58 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 multiview pseudo-labeling (ablation)Pseudo-Label · UCF101 split-1 10% labelstop-1 accuracy: 30.4 -> 70.0Improved
add multiview pseudo-labeling (ablation)UDA · UCF101 split-1 10% labelstop-1 accuracy: 47.0 -> 77.8Improved
add multiview pseudo-labeling (ablation)FixMatch · UCF101 split-1 10% labelstop-1 accuracy: 48.5 -> 79.1Improved
add optical flow view to RGB (ablation)FixMatch MvPL · UCF101 split-1 10% labelstop-1 accuracy: 48.5 -> 76.5Improved
add temporal gradient view to RGB (ablation)FixMatch MvPL · UCF101 split-1 10% labelstop-1 accuracy: 48.5 -> 74.0Improved
add both flow and temporal gradients to RGB (ablation)FixMatch MvPL · UCF101 split-1 10% labelstop-1 accuracy: 48.5 -> 79.1Improved
use aggregated-supervision (All) instead of self-supervision (ablation)pseudo-label generation · UCF101 split-1 10% labelstop-1 accuracy: 75.8 -> 79.1Improved
use aggregated-supervision (All) instead of random-supervision (ablation)pseudo-label generation · UCF101 split-1 10% labelstop-1 accuracy: 75.5 -> 79.1Improved
use aggregated-supervision (All) instead of excluding self view (ablation)pseudo-label generation · UCF101 split-1 10% labelstop-1 accuracy: 78.1 -> 79.1Improved
add 80 epochs supervised warm-up (ablation)MvPL · UCF101 split-1 10% labelstop-1 accuracy: 79.1 -> 80.5Improved
lengthen warm-up from 80 to 160 epochs (ablation)MvPL · UCF101 split-1 10% labelstop-1 accuracy: 80.5 -> 80.3Hurt
use MvPL instead of supervised baselineR-50 Slow · Kinetics-400 (1% labels)top-1 accuracy: 5.2 -> 17.0Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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