AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

$K$-Shot Contrastive Learning of Visual Features With Multiple Instance Augmentations

Haohang Xu, Hongkai Xiong, Guo-Jun Qi

TPAMI 2021 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning

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
replace pairwise cosine similarity with projection onto K=5 augmentation subspace (K-shot contrastive learning)MoCo v2 (ResNet-50, 200 epochs) · ImageNet linear evaltop-1 accuracy: 67.5 -> 68.8Improved
apply K-shot subspace contrastive loss over longer pretrainingMoCo v2 (rerun, 800 epochs) · ImageNet linear evaltop-1 accuracy: 70.6 -> 71.4Improved
increase number of augmentations K from 1 to 3 (ablation)KSCL, rho=40% · ImageNet linear evaltop-1 accuracy: 67.2 -> 68.5Improved
increase number of augmentations K from 3 to 5 (ablation)KSCL, rho=40% · ImageNet linear evaltop-1 accuracy: 68.5 -> 68.8Improved
keep only 40% of eigenvalue mass instead of 90% (ablation)KSCL, K=5 · ImageNet linear evaltop-1 accuracy: 68.4 -> 68.8Improved
increase K from 1 to 5 raises training cost (ablation)KSCL · ImageNet linear evaltraining time per epoch (min): 16 -> 37Hurt
K=1 KSCL (projection length) versus cosine similarity of MoCo v2 (ablation)KSCL K=1 · ImageNet linear evaltop-1 accuracy: 67.5 -> 67.2No clear effect
use KSCL pretrained backbone instead of MoCo v2 (200 epochs) for detectionFaster RCNN on VOC · VOC07 test detectionAP: 57.0 -> 57.1Improved
use KSCL pretrained backbone instead of MoCo v2 (800 epochs) for detectionFaster RCNN on VOC · VOC07 test detectionAP75: 64.0 -> 64.2Improved
use a MoCo queue-based KSCL compared with SimCLR large batch baseline gainKSCL vs MoCo v2 over SimCLR 8192 baseline · ImageNet linear evaltop-1 accuracy: 66.6 -> 68.8Improved

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