AI Paper Atlas

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

A Hyperdimensional One Place Signature to Represent Them All: Stackable Descriptors For Visual Place Recognition

Connor Malone, Somayeh Hussaini, Tobias Fischer, Michael Milford

ICCV 2025 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 5 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
fuse multiple reference sets with HDC bundling (HOPS) instead of best single referenceSALAD · Oxford RobotCar Dusk queryRecall@1: 76.3 -> 87.1Improved
fuse references with HOPS instead of best single referenceSALAD · Nordland (4 queries, avg)Recall@1: absolute 2.9%Improved
fuse references with HOPSSALAD · SFU-Mountain Dusk queryRecall@1: 99.0% to 100%Improved
use HOPS vs distance-matrix averaging and poolingdescriptors · all datasetsRecall@1: 69 out of 90 casesImproved
use HOPS with 512D descriptors CosPlace/EigenPlaces on RobotCar Night queryCosPlace/EigenPlaces · Oxford RobotCar NightRecall@1: 1.8% and 2.4% reductionHurt
use HOPS with CosPlace on Nordland Summer queryCosPlace · Nordland SummerRecall@1: 1.0% reductionHurt
use pooling vs dMat averaging for low-dimensional descriptorsCosPlace/EigenPlaces · Oxford RobotCarRecall@1: pooling better for low dimensionMixed
reduce descriptor dimensionality with Gaussian random projection while fusing with HOPS (ablation)CricaVPR · Oxford RobotCar DuskRecall@1: 10752D to 512D, 85.7No clear effect
replace real extra references with synthetic augmentations in HOPSSALAD · Oxford RobotCar DuskRecall@1: 73.6 -> 76.1Improved
replace real extra references with synthetic augmentations in HOPSSALAD · Oxford RobotCar OvercastRecall@1: 84.8 -> 84.2Hurt
fuse 7-9 references per place with HOPS instead of keeping all referencesSALAD · Google Landmarks v2 microRecall@1: 69.7 -> 65.8Hurt
use HOPS fusion instead of dimensionality reduction to 1024D at same memorySALAD · Google Landmarks v2 microRecall@1: 59.7 -> 65.8Improved

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