AI Paper Atlas

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

Towards Test-time Efficient Visual Place Recognition via Asymmetric Query Processing

Jaeyoon Kim, Yoonki Cho, Sung-Eui Yoon

AAAI 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 0 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 implicit embedding augmentation to asymmetric contrastive loss (ablation)asymmetric VPR (BoQ gallery, EfficientViT-B2 query) · NordlandRecall@1: 70.3 -> 74.6Improved
add implicit embedding augmentation (ablation)asymmetric VPR (BoQ gallery, EfficientViT-B2 query) · Tokyo24/7Recall@1: 90.2 -> 92.7Improved
use implicit instead of explicit (K=10) embedding augmentation (ablation)asymmetric VPR (BoQ gallery, EfficientViT-B2 query) · NordlandRecall@1: 73.3 -> 74.6Improved
use geographical memory bank instead of queue-based memory bank (ablation)asymmetric VPR (BoQ gallery, EfficientViT-B2 query) · NordlandRecall@1: 72.0 -> 74.6Improved
use same aggregator type as gallery in query network instead of MLP transformation (ablation)asymmetric VPR · Tokyo24/7Recall@1: 89.1 -> 92.7Improved
use geographical memory bank instead of k-NN (CSD) for compatible trainingasymmetric VPR · Gallery embedding precomputation timeminutes: 1392.76 -> 0.26Improved
use geographical memory bank instead of k-NN (CSD) for compatible trainingasymmetric VPR · training GPU memoryGB: 23.9 -> 17.5Improved
use asymmetric training with proposed method instead of symmetric lightweight modelEfficientViT-B2 query with BoQ gallery · NordlandRecall@1: 51.6 -> 74.6Improved
use proposed method versus best k-NN baseline on Pitts250k with MobileViTv2/BoQMobileViTv2 query with BoQ gallery · Pitts250kRecall@1: 93.2 -> 93.4No clear effect
replace DINOv2-B query with EfficientViT-B2 query (SALAD)SALAD query network · query FLOPsrelative FLOPs: 100% -> 7.3%Improved
use lightweight query network instead of high-capacity gallery modelBoQ asymmetric VPR · Pitts250kRecall@1: 96.6 -> 95.4Hurt

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