Research map / Video / Video segmentation and tracking
Video Geo-Localization Employing Geo-Temporal Feature Learning and GPS Trajectory Smoothing
Krishna Regmi, 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 clip triplet loss to frame triplet loss (ablation) | GTFL · BDD-GSV video geo-localization (SF) | localization error (m): 704.16 -> 524.81 | Improved |
| add GPS loss to frame and clip triplet losses (ablation) | GTFL · BDD-GSV video geo-localization (SF) | localization error (m): 524.81 -> 300.47 | Improved |
| add clip triplet loss to frame triplet loss (ablation) | GTFL · BDD-GSV video geo-localization (NY) | localization error (m): 978.51 -> 836.13 | Improved |
| add GPS loss to frame and clip triplet losses (ablation) | GTFL · BDD-GSV video geo-localization (Berkeley) | localization error (m): 819.54 -> 424.79 | Improved |
| add trajectory smoothing network | GTFL · BDD-GSV video geo-localization (SF) | localization error (m): 300.47 -> 128.94 | Improved |
| add trajectory smoothing network | GTFL · BDD-GSV video geo-localization (NY) | localization error (m): 493.43 -> 285.41 | Improved |
| add smoothing network | 2D CNNf · BDD-GSV video geo-localization (SF) | localization error (m): 2091.66 -> 1710.54 | Improved |
| fine-tune 2D CNN on the dataset | 2D CNN · BDD-GSV video geo-localization (SF) | localization error (m): 2516 -> 2091.66 | Improved |
| replace 2D CNNf frame-wise features with transformer geo-temporal features | 2D CNNf · BDD-GSV video geo-localization (SF) | localization error (m): 2091.66 -> 300.47 | Improved |
| use 3D CNN trained from scratch instead of fine-tuned 2D CNN | 2D CNNf · BDD-GSV video geo-localization (SF) | localization error (m): 2091.66 -> 4247.09 | Hurt |
| use 2 attention heads and 1 layer instead of 4 heads and 2 layers in smoothing network (ablation) | smoothing network · BDD-GSV video geo-localization (SF) | localization error (m): 128.94 -> 158.07 | Hurt |
| use 4 layers instead of 2 in smoothing network (4 heads) (ablation) | smoothing network · BDD-GSV video geo-localization (SF) | localization error (m): 128.94 -> 129.46 | No clear effect |
Closest papers by meaning
- GeoRouter: Dynamic Paradigm Routing for Worldwide Image Geolocalization (NeurIPS 2026 · similarity 0.86)
- GTR: A General, Multi-View, and Dynamic Framework for Trajectory Representation Learning (ICML 2025 · similarity 0.85)
- GeoRanker: Distance-Aware Ranking for Worldwide Image Geolocalization (NeurIPS 2025 · similarity 0.85)
- G3: An Effective and Adaptive Framework for Worldwide Geolocalization Using Large Multi-Modality Models (NeurIPS 2024 · similarity 0.84)
- Vision-Language Reasoning for Geolocalization: A Reinforcement Learning Approach (AAAI 2026 · similarity 0.83)
- Geolocation Representation from Large Language Models Are Generic Enhancers for Spatio-Temporal Learning (AAAI 2025 · similarity 0.83)
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.
