AI Paper Atlas

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

Soft Contrastive Learning for Visual Localization

Janine Thoma, Danda Pani Paudel, Luc V. Gool

NeurIPS 2020 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 14 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 hard positive/negative assignment (triplet) with soft assignment lossVGG-16 + NetVLAD · Oxford RobotCar + Pittsburgh mean at 5 mlocalization accuracy (%): 55.1 -> 60.7Improved
soft contrastive loss vs multi-similarity loss (previous SOTA)VGG-16 + NetVLAD · Oxford RobotCar + Pittsburgh mean at 5 mlocalization accuracy (%): 57.5 -> 60.7Improved
soft contrastive loss vs log-ratio loss (soft but proportional for all pairs)VGG-16 + NetVLAD · CMU Seasons mean at 5 mlocalization accuracy (%): 58.8 -> 70.4Improved
soft contrastive loss vs multi-similarity lossVGG-16 + NetVLAD · CMU Seasons mean at 5 mlocalization accuracy (%): 66.3 -> 70.4Improved
soft contrastive loss vs multi-similarity lossVGG-16 + NetVLAD · RobotCar Seasons mean at 5 mlocalization accuracy (%): 67.8 -> 68.9Improved
use sigmoid influence functions with loss (5) versus linear/tanh influence functions and a simple sum loss (ablation)soft contrastive loss · Oxford RobotCar and Pittsburghlocalization accuracyImproved

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