Research map / Detection and segmentation / CNN features
Where in the World Is This Image? Transformer-Based Geo-Localization in the Wild
Shraman Pramanick, Ewa M. Nowara, Joshua Gleason, Carlos D. Castillo, Rama Chellappa
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 |
|---|---|---|---|
| replace ResNet101 with ViT backbone (ablation) | RGB-only geo-localizer · Im2GPS | street accuracy (%): 14.3 -> 16.9 | Improved |
| replace ResNet101 with ViT backbone (ablation) | RGB-only geo-localizer · Im2GPS3k | street accuracy (%): 9.0 -> 9.9 | Improved |
| add segmentation branch fused at the last-layer CLS token (ablation) | ViT base · Im2GPS | street accuracy (%): 16.9 -> 17.6 | Improved |
| add multimodal feature fusion (sum CLS tokens after every layer) (ablation) | ViT + Seg · Im2GPS | street accuracy (%): 17.6 -> 19.0 | Improved |
| add multimodal feature fusion (sum CLS tokens after every layer) (ablation) | ViT + Seg · Im2GPS3k | street accuracy (%): 10.5 -> 11.1 | Improved |
| add scene recognition head (multi-task) (ablation) | ViT + Seg + MFF · Im2GPS | street accuracy (%): 19.0 -> 19.9 | Improved |
| add scene recognition head (multi-task) (ablation) | ViT + Seg + MFF · Im2GPS3k | street accuracy (%): 11.1 -> 11.8 | Improved |
| increase scene categories from 3 to 365 (ablation) | TransLocator · Im2GPS | street accuracy (%): 18.4 -> 19.9 | Improved |
| increase scene categories from 3 to 365 (ablation) | TransLocator · Im2GPS3k | street accuracy (%): 10.8 -> 11.8 | Improved |
| use fine geo-cells instead of coarse geo-cells (ablation) | TransLocator · Im2GPS | street accuracy (%): 9.8 -> 18.1 | Improved |
| combine coarse, middle and fine geo-cell losses instead of fine only (ablation) | TransLocator · Im2GPS | street accuracy (%): 18.1 -> 19.0 | Improved |
| add TenCrop test augmentation (ablation) | TransLocator · Im2GPS | street accuracy (%): 19.0 -> 19.9 | Improved |
Closest papers by meaning
- GeoRanker: Distance-Aware Ranking for Worldwide Image Geolocalization (NeurIPS 2025 · similarity 0.89)
- PIGEON: Predicting Image Geolocations (CVPR 2024 · similarity 0.88)
- GeoRouter: Dynamic Paradigm Routing for Worldwide Image Geolocalization (NeurIPS 2026 · similarity 0.88)
- Around the World in 80 Timesteps: A Generative Approach to Global Visual Geolocation (CVPR 2025 · similarity 0.88)
- Scaling Image Geo-Localization to Continent Level (NeurIPS 2025 · similarity 0.87)
- OpenStreetView-5M: The Many Roads to Global Visual Geolocation (CVPR 2024 · 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.
