AI Paper Atlas

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

Rethinking Preventing Class-Collapsing in Metric Learning With Margin-Based Losses

Elad Levi, Tete Xiao, Xiaolong Wang, Trevor Darrell

ICCV 2021 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 23 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
replace random positive sampling with easy positive sampling (EPS)Triplet loss · MNIST test digits (unseen)Recall@1: 35.2 -> 42.3Improved
add EPS to margin loss with distance samplingDist-Margin · Cars-196Recall@1: 79.6 -> 83.6Improved
add EPS to margin loss with distance samplingDist-Margin · CUB-200-2011Recall@1: 63.6 -> 64.7Improved
add EPS to multi-similarity lossMS · Cars-196Recall@1: 81.2 -> 82.9Improved
add EPS to multi-similarity lossMS · CUB-200-2011Recall@1: 62.3 -> 63.3Improved
add EPS to triplet loss with semi-hard samplingTrip. + SH · Cars-196Recall@1: 76.1 -> 78.3Improved
add EPS to triplet loss with semi-hard samplingTrip. + SH · CUB-200-2011Recall@1: 61.5 -> 61.8No clear effect
add EPS to triplet loss with semi-hard samplingTrip. + SH · Omniglot-lettersRecall@1: 49.4 -> 68.4Improved
add EPS to multi-similarity lossMS · Omniglot-languagesRecall@1: 78.8 -> 86.2Improved
add EPS to semi-hard samplingTriplet · Omniglot training set, languagesRecall@1: 99.9 -> 94.5Hurt
add EPS to semi-hard samplingTriplet · Omniglot training set, lettersRecall@1: 70.3 -> 77.5Improved
add EPS to MS with 512-d embedding and Inception backboneMS · Cars-196Recall@1: 84.1 -> 85.5Improved

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