Update README.md
Browse files
README.md
CHANGED
|
@@ -12,6 +12,16 @@ size_categories:
|
|
| 12 |
|
| 13 |
# GeoSR-Bench
|
| 14 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 15 |
GeoSR-Bench directly connects super-resolution (SR) with downstream Earth monitoring tasks, moving beyond conventional fidelity-based evaluation. It comprises spatially co-located, temporally aligned, and quality-controlled image pairs from about 36,000 locations across diverse land covers, spanning spatial resolutions from 500 m to 0.6 m. It is designed to evaluate whether improved image resolution from SR models translates into better downstream performance for tasks such as land cover segmentation, infrastructure mapping, and biophysical variable estimation.
|
| 16 |
GeoSR-Bench includes two cross-platform super-resolution tasks:
|
| 17 |
|
|
|
|
| 12 |
|
| 13 |
# GeoSR-Bench
|
| 14 |
|
| 15 |
+
## 🚧 Dataset Upload in Progress
|
| 16 |
+
|
| 17 |
+
GeoSR-Bench is currently being uploaded and reorganized.
|
| 18 |
+
|
| 19 |
+
Some files, metadata, and subsets may still be incomplete or subject to change.
|
| 20 |
+
|
| 21 |
+
---
|
| 22 |
+
|
| 23 |
+
## Dataset Description
|
| 24 |
+
|
| 25 |
GeoSR-Bench directly connects super-resolution (SR) with downstream Earth monitoring tasks, moving beyond conventional fidelity-based evaluation. It comprises spatially co-located, temporally aligned, and quality-controlled image pairs from about 36,000 locations across diverse land covers, spanning spatial resolutions from 500 m to 0.6 m. It is designed to evaluate whether improved image resolution from SR models translates into better downstream performance for tasks such as land cover segmentation, infrastructure mapping, and biophysical variable estimation.
|
| 26 |
GeoSR-Bench includes two cross-platform super-resolution tasks:
|
| 27 |
|