English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Visual place recognition has gained significant attention in recent years asa crucial technology in autonomous driving and robotics. Currently, the twomain approaches are the perspective view retrieval (P2P) paradigm and theequirectangular image retrieval (E2E) paradigm. However, it is practical andnatural to assume that users only have consumer-grade pinhole cameras to obtainquery perspective images and retrieve them in panoramic database images frommap providers. To address this, we propose \textit{PanoVPR}, aperspective-to-equirectangular (P2E) visual place recognition framework thatemploys sliding windows to eliminate feature truncation caused by hardcropping. Specifically, PanoVPR slides windows over the entire equirectangularimage and computes feature descriptors for each window, which are then comparedto determine place similarity. Notably, our unified framework enables directtransfer of the backbone from P2P methods without any modification, supportingnot only CNNs but also Transformers. To facilitate training and evaluation, wederive the Pitts250k-P2E dataset from the Pitts250k and establish YQ360, latteris the first P2E visual place recognition dataset collected by a mobile robotplatform aiming to simulate real-world task scenarios better. Extensiveexperiments demonstrate that PanoVPR achieves state-of-the-art performance andobtains 3.8% and 8.0% performance gain on Pitts250k-P2E and YQ360 compared tothe previous best method, respectively. Code and datasets will be publiclyavailable at https://github.com/zafirshi/PanoVPR.

PanoVPR: Towards Unified Perspective-to-Equirectangular Visual Place Recognition via Sliding Windows across the Panoramic View