Datasets:
ProGamerGov
commited on
Commit
·
b4dfacf
1
Parent(s):
819f3c8
Better sorting of datasets in description
Browse files
README.md
CHANGED
@@ -7,13 +7,15 @@ A collection of regularization / class instance datasets for the [Stable Diffusi
|
|
7 |
|
8 |
Currently this repository contains the following datasets (datasets are named after the prompt they used):
|
9 |
|
|
|
|
|
10 |
* "**artwork style**": 4125 images generated using 50 DDIM steps and a CFG of 7, using the MSE VAE.
|
11 |
|
12 |
* "**illustration style**": 3050 images generated using 50 DDIM steps and a CFG of 7, using the MSE VAE.
|
13 |
|
14 |
-
* "**
|
15 |
|
16 |
-
|
17 |
|
18 |
* "**person**": 2115 images generated using 50 DDIM steps and a CFG of 7, using the MSE VAE.
|
19 |
|
@@ -21,12 +23,20 @@ Currently this repository contains the following datasets (datasets are named af
|
|
21 |
|
22 |
* "**guy**": 4820 images generated using 50 DDIM steps and a CFG of 7, using the MSE VAE.
|
23 |
|
24 |
-
* "**erotic photography**": 2760 images generated using 50 DDIM steps and a CFG of 7, using the MSE VAE.
|
25 |
-
|
26 |
* "**supermodel**": 4411 images generated using 50 DDIM steps and a CFG of 7, using the MSE VAE.
|
27 |
|
|
|
|
|
|
|
|
|
28 |
* "**kitty**": 5100 images generated using 50 DDIM steps and a CFG of 7, using the MSE VAE.
|
29 |
|
|
|
|
|
|
|
|
|
|
|
|
|
30 |
|
31 |
I used the "Generate Forever" feature in [AUTOMATIC1111's WebUI](https://github.com/AUTOMATIC1111/stable-diffusion-webui) to create thousands of images for each dataset. Every image in a particular dataset uses the exact same settings, with only the seed value being different.
|
32 |
|
|
|
7 |
|
8 |
Currently this repository contains the following datasets (datasets are named after the prompt they used):
|
9 |
|
10 |
+
Art Styles
|
11 |
+
|
12 |
* "**artwork style**": 4125 images generated using 50 DDIM steps and a CFG of 7, using the MSE VAE.
|
13 |
|
14 |
* "**illustration style**": 3050 images generated using 50 DDIM steps and a CFG of 7, using the MSE VAE.
|
15 |
|
16 |
+
* "**erotic photography**": 2760 images generated using 50 DDIM steps and a CFG of 7, using the MSE VAE.
|
17 |
|
18 |
+
People
|
19 |
|
20 |
* "**person**": 2115 images generated using 50 DDIM steps and a CFG of 7, using the MSE VAE.
|
21 |
|
|
|
23 |
|
24 |
* "**guy**": 4820 images generated using 50 DDIM steps and a CFG of 7, using the MSE VAE.
|
25 |
|
|
|
|
|
26 |
* "**supermodel**": 4411 images generated using 50 DDIM steps and a CFG of 7, using the MSE VAE.
|
27 |
|
28 |
+
* "**bikini model**": 4260 images generated using 50 DDIM steps and a CFG of 7, using the MSE VAE.
|
29 |
+
|
30 |
+
Animals
|
31 |
+
|
32 |
* "**kitty**": 5100 images generated using 50 DDIM steps and a CFG of 7, using the MSE VAE.
|
33 |
|
34 |
+
Vehicles
|
35 |
+
|
36 |
+
* "**fighter jet**": 1600 images generated using 50 DDIM steps and a CFG of 7, using the MSE VAE.
|
37 |
+
|
38 |
+
* "**train**": 2669 images generated using 50 DDIM steps and a CFG of 7, using the MSE VAE.
|
39 |
+
|
40 |
|
41 |
I used the "Generate Forever" feature in [AUTOMATIC1111's WebUI](https://github.com/AUTOMATIC1111/stable-diffusion-webui) to create thousands of images for each dataset. Every image in a particular dataset uses the exact same settings, with only the seed value being different.
|
42 |
|