AI Paper Atlas

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

It Takes Two to Tango: Mixup for Deep Metric Learning

Shashanka Venkataramanan, Bill Psomas, Ewa Kijak, laurent amsaleg, Konstantinos Karantzalos, Yannis Avrithis

ICLR 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Data augmentation and generalization bias · 9 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 Metrix feature mixup with interpolated labelsMulti-Similarity · CUB200Recall@1: 67.8 -> 71.4Improved
Add Metrix feature mixup with interpolated labelsMulti-Similarity · Cars196Recall@1: 87.8 -> 89.6Improved
Add Metrix feature mixup with interpolated labelsMulti-Similarity · SOPRecall@1: 76.9 -> 81.0Improved
Add Metrix feature mixup with interpolated labelsMulti-Similarity · In-ShopRecall@1: 90.1 -> 92.2Improved
Use feature-level mixup instead of input or embedding mixup (ablation)Contrastive loss · Cars196Recall@1: 84.7 vs 85.1Improved
Use all pairs rather than 20 or 40 hard negatives for feature mixup (ablation)Metrix feature mixup · Cars196Recall@1: 83.5, 84.0 -> 85.1Improved
Increase input mixup hard negatives from k=1 to k=3 (ablation)Metrix input mixup · Cars196Recall@1: 82.0 -> 82.9Improved
Mix positive-negative pairs instead of positive-positive pairs (ablation)Metrix feature mixup · Cars196Recall@1: 81.1 -> 84.0Improved
Mix anchor-negative pairs instead of positive-positive pairs (ablation)Metrix feature mixup · Cars196Recall@1: 81.1 -> 83.7Improved
Mix positive-positive pairs only (ablation)Contrastive loss · Cars196Recall@1: 81.6 -> 81.1Hurt
Combine input, feature, and embedding mixup (ablation)Metrix · Cars196Recall@1: 85.1 -> 85.3Improved
Combine feature and embedding mixup (ablation)Metrix · Cars196Recall@1: 85.1 -> 84.7Hurt

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