Datasets:
calm-and-collected
commited on
Commit
•
b70e176
1
Parent(s):
0b615d5
Update README.md
Browse files
README.md
CHANGED
@@ -29,35 +29,30 @@ The dataset is strucured as follows:
|
|
29 |
|---|---|---|---|
|
30 |
| 646 | 1.6 Gb | Yes | No |
|
31 |
|
|
|
|
|
|
|
|
|
|
|
32 |
|
|
|
|
|
|
|
|
|
33 |
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
During manual annotation the following features were consitently annotated:
|
48 |
-
- Type of Postcard (drawing, photograph, colored in photograph)
|
49 |
-
- Aspect ration (horizontal, vertical or square)
|
50 |
-
- Border color (if there is a border)
|
51 |
-
- Damage of the postcard (ranging from no annotation, slightly damage, damage, significant damage)
|
52 |
-
- Stamps
|
53 |
-
- Folding damage
|
54 |
-
- Lineart
|
55 |
-
- Monochrome (color images are not specified)
|
56 |
-
|
57 |
-
Image dataset composition:
|
58 |
-
The dataset compromises of postcards originating from Germany, Poland, Russia and the United States of America. No additional annotation provded to identify where the postcards are from.
|
59 |
-
Most of the postcards depict a bias towards nature scenes E.G snowy mountain valleys at sunset. Training a model could create a bias towards these images.
|
60 |
-
</details>
|
61 |
|
62 |
## license
|
63 |
This dataset is licensed under CC BY 4.0 Deed. This gives you the rights to:
|
|
|
29 |
|---|---|---|---|
|
30 |
| 646 | 1.6 Gb | Yes | No |
|
31 |
|
32 |
+
### Collection method:
|
33 |
+
Manual search of WikiMedia pages and selection of images with attributes that allow for the usage of images without permission or attribution of the creator of the media. Licenses include:
|
34 |
+
- CC-O
|
35 |
+
- No license
|
36 |
+
- Public domain
|
37 |
|
38 |
+
### Annotation method:
|
39 |
+
The data was annotated using Kohya_SS in 2 phases:
|
40 |
+
1. Automated annotation using Clip.
|
41 |
+
2. Manual annotation.
|
42 |
|
43 |
+
During manual annotation the following features were consitently annotated:
|
44 |
+
- Type of Postcard (drawing, photograph, colored in photograph)
|
45 |
+
- Aspect ration (horizontal, vertical or square)
|
46 |
+
- Border color (if there is a border)
|
47 |
+
- Damage of the postcard (ranging from no annotation, slightly damage, damage, significant damage)
|
48 |
+
- Stamps
|
49 |
+
- Folding damage
|
50 |
+
- Lineart
|
51 |
+
- Monochrome (color images are not specified)
|
52 |
+
|
53 |
+
### Image dataset composition:
|
54 |
+
The dataset compromises of postcards originating from Germany, Poland, Russia and the United States of America. No additional annotation provded to identify where the postcards are from.
|
55 |
+
Most of the postcards depict a bias towards nature scenes E.G snowy mountain valleys at sunset. Training a model could create a bias towards these images.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
56 |
|
57 |
## license
|
58 |
This dataset is licensed under CC BY 4.0 Deed. This gives you the rights to:
|