Update README.md
Browse files
README.md
CHANGED
|
@@ -7,7 +7,7 @@ license:
|
|
| 7 |
multilinguality: []
|
| 8 |
pretty_name: Full Body Anime HQ
|
| 9 |
size_categories:
|
| 10 |
-
-
|
| 11 |
source_datasets:
|
| 12 |
- original
|
| 13 |
tags: []
|
|
@@ -22,12 +22,13 @@ FBAnimeHQ is a dataset with high-quality full-body anime girl images in a resolu
|
|
| 22 |
|
| 23 |
### Dataset Summary
|
| 24 |
|
| 25 |
-
The dataset contains
|
| 26 |
|
| 27 |
All images are on white background
|
| 28 |
|
| 29 |
### Collection Method
|
| 30 |
|
|
|
|
| 31 |
Collect from danbooru website.
|
| 32 |
|
| 33 |
Use yolov5 to detect and clip image.
|
|
@@ -38,6 +39,10 @@ Use deepdanbooru to filter image.
|
|
| 38 |
|
| 39 |
Finally clean the dataset manually.
|
| 40 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 41 |
### Contributions
|
| 42 |
|
| 43 |
Thanks to [@SkyTNT](https://github.com/SkyTNT) for adding this dataset.
|
|
|
|
| 7 |
multilinguality: []
|
| 8 |
pretty_name: Full Body Anime HQ
|
| 9 |
size_categories:
|
| 10 |
+
- 100K<n<1M
|
| 11 |
source_datasets:
|
| 12 |
- original
|
| 13 |
tags: []
|
|
|
|
| 22 |
|
| 23 |
### Dataset Summary
|
| 24 |
|
| 25 |
+
The dataset contains 112,806 images.
|
| 26 |
|
| 27 |
All images are on white background
|
| 28 |
|
| 29 |
### Collection Method
|
| 30 |
|
| 31 |
+
#### v1.0
|
| 32 |
Collect from danbooru website.
|
| 33 |
|
| 34 |
Use yolov5 to detect and clip image.
|
|
|
|
| 39 |
|
| 40 |
Finally clean the dataset manually.
|
| 41 |
|
| 42 |
+
#### v2.0
|
| 43 |
+
|
| 44 |
+
Base on v1.0, use Novelai image-to-image to enhance and expand the dataset.
|
| 45 |
+
|
| 46 |
### Contributions
|
| 47 |
|
| 48 |
Thanks to [@SkyTNT](https://github.com/SkyTNT) for adding this dataset.
|