RoofD is dataset for flat roof defect segmentation
RoofD dataset contains 6400 color image pairs (aerial image and segmentation mask) with size of 1024x576 divided on two parts: train (5184 image pairs) and test (1216 image pairs).
Roof defect segment catergories:
“hollows” (a), “swellings” (b), “folds” (c), “patches” (d) and “breaks” (e)
Details of RoofD Dataset:
| Segment category | Train ID | Color | Pixels (train) | Pixels (test) | Images (train) | Images (test) |
|---|---|---|---|---|---|---|
| background | 0 | (0, 0, 0) | 2749885350 | 638910979 | 5184 | 1216 |
| hollows | 1 | (255, 255, 0) | 154395478 | 34493513 | 3598 | 751 |
| swellings | 2 | (153, 255, 204) | 143853128 | 42520381 | 3307 | 896 |
| folds | 3 | (6, 89, 182) | 8763604 | 1098290 | 1199 | 208 |
| patches | 4 | (0, 255, 255) | 642958 | 126472 | 107 | 31 |
| breaks | 5 | (255, 193, 255) | 107098 | 76349 | 54 | 32 |
Download RoofD archive (5.7 GB) which contains:
- RoofD train sample (3.7 GB, 5184 image pairs)
- RoofD test sample (0.9 GB, 1216 image pairs)