MegaLoc Image Retrieval Model Achieves SOTA Across Multiple VPR Benchmarks

rsasaki0109 · x · 2026-07-31

MegaLoc is a new universal image retrieval model that achieves State-of-the-Art (SOTA) performance on most Visual Place Recognition (VPR) datasets, covering both indoor and outdoor scenarios.

The project features a Gradio demo utilizing a database of roughly 5 million street-view images from San Francisco for similarity matching. Developers can easily load and integrate the model via torch.hub.

Original post →

More from Research

Research channel →