AI Paper Atlas

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

Self-supervising Fine-grained Region Similarities for Large-scale Image Localization

Yixiao Ge, Haibo Wang, Feng Zhu, Rui Zhao, Hongsheng Li

ECCV 2020 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 179 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 top-k difficult positives with hard labels (ablation)Baseline · Tokyo 24/7recall R@1 (%): 80.6 -> 76.2Hurt
add soft self-enhanced similarity loss to top-k positives training (ablation)Baseline w/ top-k positives · Tokyo 24/7recall R@1 (%): 76.1% to 80.6%Improved
add region augmentation with GPS labels only (ablation)Baseline · Tokyo 24/7recall R@1 (%): 80.6 -> 79.8Hurt
add all image-to-region similarities (half and quarter regions) (ablation)Ours · Tokyo 24/7recall R@1 (%): 80.6 -> 85.4Improved
add quarter regions to half regions (ablation)Ours · Tokyo 24/7recall R@1 (%): 84.4 -> 85.4Improved
add positive sub-regions (ablation)Ours · Tokyo 24/7recall R@1 (%): 80.8 -> 85.4Improved
add negative sub-regions mining (ablation)Ours · Tokyo 24/7recall R@1 (%): 82.2 -> 85.4Improved
add k-reciprocal nearest neighbor top-k ranking refinery (ablation)Ours · Tokyo 24/7recall R@1 (%): 83.8 -> 85.4Improved
add softmax temperature annealing across generations (ablation)Ours · Tokyo 24/7recall R@1 (%): 83.5 -> 85.4Improved
train more generations (1st to 4th)Ours · Tokyo 24/7recall R@1 (%): 80.6 -> 82.6 -> 84.2 -> 85.4Improved
replace SARE with proposed methodSARE · Tokyo 24/7recall R@1 (%): 79.7 -> 85.4Improved
replace SARE with proposed methodSARE · Pitts250k-testrecall R@1 (%): 89.0 -> 90.7Improved

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