xBD
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Building Extraction | xBD pre | IoU43.1 | 24 | |
| Active Geo-localization | xBD-disaster zero-shot generalization 10x10 grid | Success Ratio26.5 | 15 | |
| Building Damage Assessment | xBD (test) | F1 Score (Overall)80.6 | 15 | |
| Building Damage Assessment | xBD v1 (test) | F1 (Floc)0.8738 | 14 | |
| Vision-Language Captioning | xBD (test) | CLIPScore (%)77.86 | 8 | |
| Building Detection | xBD (test) | F1 Score (Building)88.43 | 8 | |
| Building Damage Detection | xBD xView2 (test) | F1 Score63.8 | 7 | |
| Aerial Goal Localization | xBD-disaster (test) | SR (C=4)40.02 | 5 | |
| Aerial Goal Localization | xBD pre (test) | Success Ratio (C=4)38.25 | 5 | |
| Building Damage Assessment | xBD xView2 (holdout) | F1 Overall77.02 | 5 | |
| Building Damage Assessment | xBD | F1loc86.86 | 4 | |
| Damage Assessment | xBD Hurricane Matthew upscaled full image | VLM-As-A-Jury90.22 | 4 | |
| Damage Assessment | xBD Moore Tornado upscaled full image | VLM-As-A-Jury93.93 | 4 | |
| Damage Assessment | xBD Matthew Hurricane upscaled full images v1.0 | Average CLIPScore62.42 | 4 | |
| Damage Assessment | xBD Moore Tornado upscaled full images v1.0 | Average CLIPScore63.34 | 4 | |
| Building Damage Assessment | xBD Hurricane Matthew (cropped building images, upscaled) | Average CLIPScore58.11 | 4 | |
| Building Damage Assessment | xBD Moore Tornado (cropped building images, upscaled) | Average CLIPScore59.66 | 4 | |
| Damage Assessment | xBD Matthew Hurricane cropped buildings | VLM-As-A-Jury87.64 | 4 | |
| Damage Assessment | xBD Moore Tornado subset cropped buildings | VLM-As-A-Jury88.6 | 4 | |
| Scene Classification | xBD Temporal Set 1 | Accuracy96.37 | 4 | |
| Damage Assessment | xBD general upscaled full images v1.0 | Average CLIPScore60.6 | 2 |
Showing 21 of 21 rows