AI Paper Atlas

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

Sharing Matters for Generalization in Deep Metric Learning

Timo Milbich, Karsten Roth, Biagio Brattoli, Björn Ommer

TPAMI 2020 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 28 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 shared-feature task with separate encoderMargin loss ResNet50 · CARS196Recall@1: 81.5 -> 87.0Improved
add shared-feature taskMargin loss ResNet50 · CUB200-2011Recall@1: 65.2 -> 68.6Improved
add shared-feature taskMargin loss ResNet50 · SOPRecall@1: 76.1 -> 78.2Improved
add shared-feature taskSemihard triplet loss ResNet50 · CARS196Recall@1: 72.7 -> 79.4Improved
add shared-feature taskN-pairs ResNet50 · CARS196Recall@1: 70.6 -> 77.2Improved
add shared-feature taskProxyNCA ResNet50 · CARS196Recall@1: 80.8 -> 82.7Improved
add shared task to discriminative baseline with single encoder (ablation)discr. only · CARS196Recall@1: 81.5 -> 83.2Improved
use separate encoders for discriminative and shared tasks (ablation)both · CARS196Recall@1: 83.2 -> 84.1Improved
add decorrelation between embeddings (ablation)both+sep · CARS196Recall@1: 84.1 -> 87.0Improved
increase decorrelation weight too large (ablation)both+sep+decor · CARS196Recall@1: 87.0 -> 82.1Hurt
train only on shared triplets without discriminative task (ablation)discr. only · CARS196Recall@1: 81.5 -> 36.1Hurt
add feature standardization before grouping (ablation)group triplets · CARS196Recall@1: 82.3 -> 85.8Improved

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