AI Paper Atlas

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

Do Different Deep Metric Learning Losses Lead to Similar Learned Features?

Konstantin Kobs, Michael Steininger, Andrzej Dulny, Andreas Hotho

ICCV 2021 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 8 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
compare loss pairs of different type (ranking vs classification) instead of same typeDML saliency maps · SOPsaliency correlation (%): 62 -> 46Hurt
compare ranking with classification losses instead of same typeDML saliency maps · Cars196saliency correlation (%): 86 -> 85No clear effect
compare ranking with classification losses instead of same typeDML saliency maps · CUB200saliency correlation (%): 90 -> 90No clear effect
compare trained model against untrained None modelDML saliency maps · Cars196saliency correlation (%): 86 -> 75Hurt
train with DML loss rather than ImageNet init only (None)car model clustering · Rendered cars: car modelNR-Prec: 50.61 -> 57.91Improved
train with DML loss rather than ImageNet init only (None)background color value clustering · Rendered cars: background color valueNR-Prec: 6.80 -> 1.62Improved
use classification instead of ranking losscar color hue influence · Rendered cars: car color hueNR-Prec: 3.25 -> 4.07Hurt
use classification instead of ranking lossbackground color saturation influence · Rendered cars: background color saturationNR-Prec: 1.48 -> 2.22Hurt
use classification instead of ranking losssun elevation influence · Rendered cars: sun elevationNR-Prec: 8.29 -> 8.94Hurt
use classification instead of ranking losscar rotation influence · Rendered cars: car rotationNR-Prec: 39.07 -> 37.84Improved
use classification instead of ranking losscamera height influence · Rendered cars: camera heightNR-Prec: 20.73 -> 19.35Improved
use classification instead of ranking losscar model clustering · Rendered cars: car modelNR-Prec: 57.91 -> 56.89No clear effect

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