AI Paper Atlas

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

Seed the Views: Hierarchical Semantic Alignment for Contrastive Representation Learning

Haohang Xu, Xiaopeng Zhang, Hao Li, Lingxi Xie, Wenrui Dai, Hongkai Xiong, Qi Tian

TPAMI 2022 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Empirical contrastive visual learning · 48 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 one KNN-selected positive sample (ablation)MoCo v2 · ImageNet linear eval (200 epochs)top-1 accuracy: 67.5 -> 70.4Improved
add CutMix mixed sample as extra positive (ablation)CsMl · ImageNet linear eval (200 epochs)top-1 accuracy: 70.4 -> 71.6Improved
add multi-crop augmentation (ablation)CsMl · ImageNet linear eval (200 epochs)top-1 accuracy: 71.6 -> 74.6Improved
replace CutMix with MixUp for the mixed sample (ablation)CsMl · ImageNet linear eval (200 epochs)top-1 accuracy: 71.6 -> 70.1Hurt
add a second KNN positive instead of mixed sample (ablation)CsMl · ImageNet linear eval (200 epochs)top-1 accuracy: 70.4 -> 70.2No clear effect
use identical sample as positive (k=0) instead of KNN neighbors (ablation)MoCo v2 · ImageNet linear eval (200 epochs)top-1 accuracy: 67.5 -> 68.6Hurt
add multi-level pulling at stage 3 and 4 (ablation)CsMl · ImageNet linear eval stage 3top-1 accuracy: 60.7 -> 64.0Improved
add pulling at stage 2, 3 and 4 versus MoCo v2 (ablation)CsMl · ImageNet 1% labels fine-tuningtop-1 accuracy: 52.4 -> 54.8Improved
add multi-level pulling (ablation)CsMl · COCO detectionAP bb: 39.5 -> 40.3Improved
apply CsMl to BYOLBYOL · ImageNet linear eval (300 epochs)top-1 accuracy: 72.5 -> 75.3Improved
add data mining with pseudo labelsCsMl · ImageNet 1% labelstop-1 accuracy: 62.2 -> 66.7Improved
compare to MoCo v2 trained for double epochs (matching samples) (ablation)CsMl · ImageNet linear evaltop-1 accuracy: 67.5 -> 68.5Improved

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