AI Paper Atlas

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

SAGE: Spatial-visual Adaptive Graph Exploration for Efficient Visual Place Recognition

Shunpeng Chen, Changwei Wang, Rongtao Xu, Peixingtian, yukun Song, Jinzhou Lin, Wenhao Xu, jingyizhang, Li Guo, Shibiao Xu

ICLR 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 1 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 Soft Probing in place of CFP aggregation (ablation)SAGE-B (OGC+GWS) · NordlandRecall@1: 95.4 -> 95.8Improved
add Soft Probing in place of CFP aggregation (ablation)SAGE-B (OGC+GWS) · SPEDRecall@1: 97.5 -> 98.0Improved
add Online Graph Creation to SoftP+GWS (ablation)SAGE-B · NordlandRecall@1: 94.2 -> 95.8Improved
add Greedy Weighted Sampling to SoftP+OGC (ablation)SAGE-B · NordlandRecall@1: 95.2 -> 95.8Improved
add Greedy Weighted Sampling to SoftP+OGC (ablation)SAGE-B · MSLS-valRecall@1: 93.6 -> 94.3Improved
online graph rebuilt each epoch vs offline mining (ablation)SAGE · SPEDRecall@1: 98.5 -> 98.9Improved
online graph rebuilt each epoch vs offline mining (ablation)SAGE · MSLS-valRecall@1: 94.2 -> 94.5Improved
use GWS instead of Cliquemining (ablation)SAGE (online) · SPEDRecall@1: 90.0 -> 98.9Improved
dynamic sampling vs Cliquemining, early training (ablation)SAGE · MSLS-val epoch 4Recall@1: 92.7 -> 93.4Improved
enlarge InteractHead feed-forward dim from 1024 to 1536 (ablation)SAGE · SPED / Pitts30kRecall@1: SPED 98.9 -> 98.6; Pitts30k 95.8 -> 96.0Mixed
add InteractHead cross-image encoder (ablation)SAGE-B · SPEDRecall@1: 91.8 -> 98.9Improved
fine-tune with frozen backbone and DPNSAGE · trainable parameters (M)trainable params: 29.8 -> 1.96Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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