Datasets:
Update README.md
Browse files
README.md
CHANGED
@@ -22,7 +22,7 @@ task_categories:
|
|
22 |
This dataset contains optical and SAR images from Sentinel-2 and Sentinel-1.
|
23 |
|
24 |
### Optical image
|
25 |
-
|
26 |
### SAR image
|
27 |
The SAR images were acquired from the Sentinel-1 satellite. Specifically, the Interferometric Wide Swath (IW) mode with a pixel size of 10 meters was used, incorporating both Vertical-Vertical (VV) and Vertical-Horizontal (VH) polarizations.The Sentinel-1 data underwent preprocessing in SNAP software, including orbit correction, thermal noise removal, radiometric calibration, and terrain correction. Finally, the data were resampled to a resolution of 10 m Γ 10 m to align with the selected Sentinel-2 spectral bands.
|
28 |
|
@@ -32,43 +32,30 @@ The SAR images were acquired from the Sentinel-1 satellite. Specifically, the In
|
|
32 |
|
33 |
| Property | Description |
|
34 |
|----------------------|-----------------------------------------|
|
35 |
-
| Source |
|
36 |
-
|
|
37 |
-
|
|
38 |
-
| Image
|
39 |
-
| Total Image Pairs | 683 |
|
40 |
-
| Preprocessing | Georeferenced and Cropped |
|
41 |
-
|
42 |
-
---
|
43 |
-
|
44 |
-
## Data Construction and Patch Sampling
|
45 |
|
46 |
-
To construct a high-quality glacier change detection dataset, we first collected bi-temporal Landsat-8 imagery covering polar regions, with an original size of 8541 Γ 8601 pixels and a spatial resolution of 30 meters. To focus on glacier-dominated areas and exclude irrelevant background, we manually selected and cropped a subregion of 5109 Γ 4801 pixels. A sliding window approach with a patch size of 256 Γ 256 pixels and a stride of 128 pixels was then applied, resulting in 683 bi-temporal image pairs. This number is determined by the spatial extent of the region and the patch extraction strategy.
|
47 |
-
|
48 |
-
The choice of 256Γ256 resolution was a trade-off between spatial detail and computational efficiency. Larger patches (e.g., 512Γ512) were avoided because they require significantly more GPU memory, reduce training speed, and tend to include a higher proportion of unchanged areas. This can lead to class imbalance, increasing the number of negative samples and causing the model to favor βno-changeβ predictions, ultimately lowering recall rate. In contrast, 256-sized patches ensure sufficient contextual information while maintaining a more balanced representation of change and no-change areas, thereby supporting more effective learning.
|
49 |
-
|
50 |
-
---
|
51 |
|
52 |
## Dataset Structure and File Organization
|
53 |
|
54 |
-
The dataset is organized into paired folders for
|
55 |
```
|
56 |
-
-Glacier-Dataset/
|
57 |
-
βββ
|
58 |
-
β βββ
|
59 |
-
β βββ
|
60 |
β βββ ...
|
61 |
-
βββ
|
62 |
-
β βββ
|
63 |
-
β βββ
|
|
|
|
|
64 |
β βββ ...
|
65 |
βββ labels/
|
66 |
-
β βββ
|
67 |
-
β βββ
|
68 |
-
β βββ ...
|
69 |
-
βββ samples/
|
70 |
-
β βββ greenland_A.png
|
71 |
-
β βββ greenland_B.png
|
72 |
β βββ ...
|
73 |
βββ README.md
|
74 |
```
|
|
|
22 |
This dataset contains optical and SAR images from Sentinel-2 and Sentinel-1.
|
23 |
|
24 |
### Optical image
|
25 |
+
The Optical images were acquired from the Sentinel-2 satellite.
|
26 |
### SAR image
|
27 |
The SAR images were acquired from the Sentinel-1 satellite. Specifically, the Interferometric Wide Swath (IW) mode with a pixel size of 10 meters was used, incorporating both Vertical-Vertical (VV) and Vertical-Horizontal (VH) polarizations.The Sentinel-1 data underwent preprocessing in SNAP software, including orbit correction, thermal noise removal, radiometric calibration, and terrain correction. Finally, the data were resampled to a resolution of 10 m Γ 10 m to align with the selected Sentinel-2 spectral bands.
|
28 |
|
|
|
32 |
|
33 |
| Property | Description |
|
34 |
|----------------------|-----------------------------------------|
|
35 |
+
| Source | Sentinel-1 SatelliteγSentinel-2 satellite|
|
36 |
+
| Spatial Resolution | 10 meters |
|
37 |
+
| Image Patch Size | 224 Γ 224 pixels |
|
38 |
+
| Total Image Pairs | 1286 |
|
|
|
|
|
|
|
|
|
|
|
|
|
39 |
|
|
|
|
|
|
|
|
|
|
|
40 |
|
41 |
## Dataset Structure and File Organization
|
42 |
|
43 |
+
The dataset is organized into paired folders for image patches and corresponding labels:
|
44 |
```
|
45 |
+
-Glacier-Dataset/LandSea
|
46 |
+
βββ opts/ (optical image)
|
47 |
+
β βββ 1_1.tif
|
48 |
+
β βββ 1_2.tif
|
49 |
β βββ ...
|
50 |
+
βββ sars/ (SAR image)
|
51 |
+
β βββ 1_1_0.tif
|
52 |
+
β βββ 1_1_1.tif
|
53 |
+
β βββ 1_2_0.tif
|
54 |
+
β βββ 1_2_1.tif
|
55 |
β βββ ...
|
56 |
βββ labels/
|
57 |
+
β βββ 1_1.tif
|
58 |
+
β βββ 1_2.tif
|
|
|
|
|
|
|
|
|
59 |
β βββ ...
|
60 |
βββ README.md
|
61 |
```
|