Add task category, link to paper
#1
by
nielsr
HF Staff
- opened
README.md
CHANGED
@@ -1,90 +1,128 @@
|
|
1 |
---
|
2 |
license: apache-2.0
|
3 |
-
|
|
|
4 |
dataset_info:
|
5 |
features:
|
6 |
-
|
7 |
-
|
8 |
-
|
9 |
-
|
10 |
-
|
11 |
-
|
12 |
-
|
13 |
-
|
14 |
-
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
|
59 |
-
|
60 |
-
|
61 |
-
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
-
|
66 |
-
|
67 |
-
|
68 |
-
|
69 |
-
|
70 |
-
|
71 |
-
|
72 |
-
|
73 |
-
|
74 |
-
|
75 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
76 |
config_name: default
|
77 |
splits:
|
78 |
-
|
79 |
-
|
80 |
---
|
|
|
81 |
# VisionRewardDB-Image
|
82 |
|
83 |
## Introduction
|
84 |
|
85 |
VisionRewardDB-Image is a comprehensive dataset designed to train VisionReward-Image models, providing detailed aesthetic annotations across 18 aspects. The dataset aims to enhance the assessment and understanding of visual aesthetics and quality. πβ¨
|
86 |
|
87 |
-
For more detail, please refer to the [**Github Repository**](https://github.com/THUDM/VisionReward). ππ
|
88 |
|
89 |
|
90 |
## Annotation Detail
|
|
|
1 |
---
|
2 |
license: apache-2.0
|
3 |
+
task_categories:
|
4 |
+
- image-text-to-text
|
5 |
dataset_info:
|
6 |
features:
|
7 |
+
- name: image
|
8 |
+
dtype: image
|
9 |
+
- name: internal_id
|
10 |
+
dtype: string
|
11 |
+
- name: prompt
|
12 |
+
dtype: string
|
13 |
+
- name: url
|
14 |
+
dtype: string
|
15 |
+
- name: annotation
|
16 |
+
struct:
|
17 |
+
- name: symmetry
|
18 |
+
dtype: int64
|
19 |
+
range:
|
20 |
+
- -1
|
21 |
+
- 1
|
22 |
+
- name: richness
|
23 |
+
dtype: int64
|
24 |
+
range:
|
25 |
+
- -2
|
26 |
+
- 2
|
27 |
+
- name: color aesthetic
|
28 |
+
dtype: int64
|
29 |
+
range:
|
30 |
+
- -1
|
31 |
+
- 1
|
32 |
+
- name: detail realism
|
33 |
+
dtype: int64
|
34 |
+
range:
|
35 |
+
- -3
|
36 |
+
- 1
|
37 |
+
- name: safety
|
38 |
+
dtype: int64
|
39 |
+
range:
|
40 |
+
- -3
|
41 |
+
- 1
|
42 |
+
- name: body
|
43 |
+
dtype: int64
|
44 |
+
range:
|
45 |
+
- -4
|
46 |
+
- 1
|
47 |
+
- name: lighting aesthetic
|
48 |
+
dtype: int64
|
49 |
+
range:
|
50 |
+
- -1
|
51 |
+
- 2
|
52 |
+
- name: lighting distinction
|
53 |
+
dtype: int64
|
54 |
+
range:
|
55 |
+
- -1
|
56 |
+
- 2
|
57 |
+
- name: background
|
58 |
+
dtype: int64
|
59 |
+
range:
|
60 |
+
- -1
|
61 |
+
- 2
|
62 |
+
- name: emotion
|
63 |
+
dtype: int64
|
64 |
+
range:
|
65 |
+
- -2
|
66 |
+
- 2
|
67 |
+
- name: main object
|
68 |
+
dtype: int64
|
69 |
+
range:
|
70 |
+
- -1
|
71 |
+
- 1
|
72 |
+
- name: color brightness
|
73 |
+
dtype: int64
|
74 |
+
range:
|
75 |
+
- -1
|
76 |
+
- 1
|
77 |
+
- name: face
|
78 |
+
dtype: int64
|
79 |
+
range:
|
80 |
+
- -3
|
81 |
+
- 2
|
82 |
+
- name: hands
|
83 |
+
dtype: int64
|
84 |
+
range:
|
85 |
+
- -4
|
86 |
+
- 1
|
87 |
+
- name: clarity
|
88 |
+
dtype: int64
|
89 |
+
range:
|
90 |
+
- -2
|
91 |
+
- 2
|
92 |
+
- name: detail refinement
|
93 |
+
dtype: int64
|
94 |
+
range:
|
95 |
+
- -4
|
96 |
+
- 2
|
97 |
+
- name: unsafe type
|
98 |
+
dtype: int64
|
99 |
+
range:
|
100 |
+
- 0
|
101 |
+
- 3
|
102 |
+
- name: object pairing
|
103 |
+
dtype: int64
|
104 |
+
range:
|
105 |
+
- -1
|
106 |
+
- 1
|
107 |
+
- name: meta_result
|
108 |
+
sequence:
|
109 |
+
dtype: int64
|
110 |
+
- name: meta_mask
|
111 |
+
sequence:
|
112 |
+
dtype: int64
|
113 |
config_name: default
|
114 |
splits:
|
115 |
+
- name: train
|
116 |
+
num_examples: 40743
|
117 |
---
|
118 |
+
|
119 |
# VisionRewardDB-Image
|
120 |
|
121 |
## Introduction
|
122 |
|
123 |
VisionRewardDB-Image is a comprehensive dataset designed to train VisionReward-Image models, providing detailed aesthetic annotations across 18 aspects. The dataset aims to enhance the assessment and understanding of visual aesthetics and quality. πβ¨
|
124 |
|
125 |
+
For more detail, please refer to the [**Github Repository**](https://github.com/THUDM/VisionReward) and the [**paper**](https://huggingface.co/papers/2412.21059). ππ
|
126 |
|
127 |
|
128 |
## Annotation Detail
|