Research map / Video / Video-language understanding
VidTAG: Temporally Aligned Video to GPS Geolocalization with Denoising Sequence Prediction at a Global Scale
Parth Parag Kulkarni, Rohit Gupta, Prakash Chandra Chhipa, Mubarak Shah
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add TempGeo to DINOv2 backbone (ablation) | VidTAG · MSLS | accuracy at 1 km (%): 26.4 -> 30.2 | Improved |
| add TempGeo to DINOv2 backbone (ablation) | VidTAG · MSLS | DFD: 9.15 -> 3.01 | Improved |
| add CLIP to DINOv2 + TempGeo (ablation) | VidTAG · MSLS | accuracy at 1 km (%): 30.2 -> 40.1 | Improved |
| add CLIP to DINOv2 + TempGeo (ablation) | VidTAG · MSLS | DFD: 3.01 -> 7.63 | Hurt |
| add GeoRefiner (ablation) | VidTAG · MSLS | DFD: 7.63 -> 3.87 | Improved |
| add GeoRefiner (ablation) | VidTAG · MSLS | accuracy at 1 km (%): 40.1 -> 41.0 | Improved |
| use DINOv2 instead of CLIP backbone (ablation) | fine-tuned backbone · MSLS | accuracy at 1 km (%): 22.5 -> 26.4 | Improved |
| use DINOv2 rather than SigLIP as second encoder alongside CLIP (ablation) | VidTAG · MSLS | accuracy at 1 km (%): 27.8 -> 40.1 | Improved |
| finer gallery resolution (1 km to 0.1 km) (ablation) | VidTAG · MSLS | accuracy at 1 km (%): 38.6 -> 41.0 | Improved |
| longer training sequences (4 to 16 frames) (ablation) | VidTAG · MSLS | accuracy at 1 km (%): 26.6 -> 33.1 | Improved |
| longer training sequences (4 to 16 frames) (ablation) | VidTAG · MSLS | median distance error (km): 2.13 km -> 1.66 km | Improved |
| alpha 100 instead of 10 in hinge loss (ablation) | VidTAG · MSLS | DFD: 3.87 -> 7.63 | Hurt |
Closest papers by meaning
- GeoRanker: Distance-Aware Ranking for Worldwide Image Geolocalization (NeurIPS 2025 · similarity 0.88)
- G3: An Effective and Adaptive Framework for Worldwide Geolocalization Using Large Multi-Modality Models (NeurIPS 2024 · similarity 0.88)
- GeoRouter: Dynamic Paradigm Routing for Worldwide Image Geolocalization (NeurIPS 2026 · similarity 0.88)
- Vision-Language Reasoning for Geolocalization: A Reinforcement Learning Approach (AAAI 2026 · similarity 0.86)
- PIGEON: Predicting Image Geolocations (CVPR 2024 · similarity 0.86)
- Around the World in 80 Timesteps: A Generative Approach to Global Visual Geolocation (CVPR 2025 · similarity 0.86)
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.
