hayden-donnelly
commited on
Commit
•
52ef48c
1
Parent(s):
f15db82
Update README.md
Browse files
README.md
CHANGED
@@ -7,4 +7,25 @@ pretty_name: Danbooru SFW 512 Filtered and Cropped Dataset
|
|
7 |
size_categories:
|
8 |
- 1M<n<10M
|
9 |
---
|
10 |
-
# Danbooru SFW 512 Filtered and Cropped
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
7 |
size_categories:
|
8 |
- 1M<n<10M
|
9 |
---
|
10 |
+
# Danbooru SFW 512 Filtered and Cropped
|
11 |
+
|
12 |
+
A version of Danbooru SFW which has been automatically filtered and cropped to 128x128 so that the resulting images focus almost entirely on characters.
|
13 |
+
First filtering was applied to remove non-character focused images, then cropping was applied to remove horizontal/vertical bars and further focus on the
|
14 |
+
central character(s) of each image. Both steps were performed automatically by two different vision models trained on manually labelled subsets of the original dataset.
|
15 |
+
This technique provided satisfactory results, however there is much room for improvement.
|
16 |
+
|
17 |
+
## Original Dataset Citation
|
18 |
+
|
19 |
+
```bibtex
|
20 |
+
@misc{danbooru2021,
|
21 |
+
author={Anonymous and Danbooru community and Gwern Branwen},
|
22 |
+
title={Danbooru2021: A Large-Scale Crowdsourced and Tagged Anime Illustration Dataset},
|
23 |
+
howpublished={\url{https://gwern.net/danbooru2021}},
|
24 |
+
url={https://gwern.net/danbooru2021},
|
25 |
+
type={dataset},
|
26 |
+
year={2022},
|
27 |
+
month={January},
|
28 |
+
timestamp={2022-01-21},
|
29 |
+
note={Accessed: 2023-12-06}
|
30 |
+
}
|
31 |
+
```
|