# VidTAG: Temporally Aligned Video to GPS Geolocalization with Denoising Sequence Prediction at a Global Scale

Source: [VidTAG: Temporally Aligned Video to GPS Geolocalization with Denoising Sequence Prediction at a Global Scale](<https://www.sota2.com/research/paper/69df1ba37d0172140ec7b9eb>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-04-16T05:00:04.684Z

Authors: Parth Parag Kulkarni, Rohit Gupta, Prakash Chandra Chhipa, Mubarak Shah

Published: 2026-04-14T00:35:07.000Z

## Abstract

The task of video geolocalization aims to determine the precise GPS coordinates of a video's origin and map its trajectory; with applications in forensics, social media, and exploration. Existing classification-based approaches operate at a coarse city-level granularity and fail to capture fine-grained details, while image retrieval methods are impractical on a global scale due to the need for extensive image galleries which are infeasible to compile. Comparatively, constructing a gallery of GPS coordinates is straightforward and inexpensive. We propose VidTAG, a dual-encoder framework that performs frame-to-GPS retrieval using both self-supervised and language-aligned features. To address temporal inconsistencies in video predictions, we introduce the TempGeo module, which aligns frame embeddings, and the GeoRefiner module, an encoder-decoder architecture that refines GPS features using the aligned frame embeddings. Evaluations on Mapillary (MSLS) and GAMa datasets demonstrate our model's ability to generate temporally consistent trajectories and outperform baselines, achieving a 20% improvement at the 1 km threshold over GeoCLIP. We also beat current State-of-the-Art by 25% on global coarse grained video geolocalization (CityGuessr68k). Our approach enables fine-grained video geolocalization and lays a strong foundation for future research. More details on the project webpage: https://parthpk.github.io/vidtag\_webpage/

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Visual Geolocalization on CityGuessr68k](<https://www.sota2.com/research/sota/visual-geolocalization-on-cityguessr68k.md>): Visual Geolocalization; CityGuessr68k; City Accuracy: 94.9.

- [Video Geolocalization on Mapillary MSLS](<https://www.sota2.com/research/sota/video-geolocalization-on-mapillary-msls.md>): Video Geolocalization; Mapillary MSLS; Success Rate @ 0.5 km: 22.2.

- [Video Geolocalization on BDD100k (GAMa)](<https://www.sota2.com/research/sota/video-geolocalization-on-bdd100k-gama.md>): Video Geolocalization; BDD100k (GAMa); Accuracy @ 0.5 km (Sub-street): 35.8.

- [Frame-wise Geolocalization on Mapillary MSLS](<https://www.sota2.com/research/sota/frame-wise-geolocalization-on-mapillary-msls.md>): Frame-wise Geolocalization; Mapillary MSLS; AR @ 0.5 km: 21.5.

- [Frame-wise Geolocalization on BDD100k (GAMa)](<https://www.sota2.com/research/sota/frame-wise-geolocalization-on-bdd100k-gama.md>): Frame-wise Geolocalization; BDD100k (GAMa); Accuracy @ 0.5 km (Sub-street): 35.4.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2604.12159>)

[PDF](<https://arxiv.org/pdf/2604.12159.pdf>)
