File size: 6,135 Bytes
3acffd5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
0483c7f
 
 
 
 
 
 
 
 
 
 
 
 
3acffd5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
0483c7f
 
 
 
 
3acffd5
 
 
 
 
 
 
 
 
 
 
 
0483c7f
 
3acffd5
 
 
 
0483c7f
3acffd5
 
0483c7f
3acffd5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
0483c7f
3acffd5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
a1bb07d
3acffd5
 
1
2
3
4
5
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
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
from datasets import load_dataset
from collections import Counter, defaultdict
import pandas as pd
from huggingface_hub import list_datasets
import os
import gradio as gr

parti_prompt_results = []
ORG = "diffusers-parti-prompts"
SUBMISSIONS = {
    "sd-v1-5": None,
    "sd-v2-1": None,
    "if-v1-0": None,
    "karlo": None,
}
LINKS = {
    "sd-v1-5": "https://huggingface.co/runwayml/stable-diffusion-v1-5",
    "sd-v2-1": "https://huggingface.co/stabilityai/stable-diffusion-2-1",
    "if-v1-0": "https://huggingface.co/DeepFloyd/IF-I-XL-v1.0",
    "karlo": "https://huggingface.co/kakaobrain/karlo-v1-alpha",
}
MODEL_KEYS = "-".join(SUBMISSIONS.keys())
SUBMISSION_ORG = f"results-{MODEL_KEYS}"

submission_names = list(SUBMISSIONS.keys())
parti_prompt_categories = load_dataset(os.path.join(ORG, "sd-v1-5"))["train"]["Category"]
parti_prompt_challenge = load_dataset(os.path.join(ORG, "sd-v1-5"))["train"]["Challenge"]


def load_submissions():
    all_datasets = list_datasets(author=SUBMISSION_ORG)
    relevant_ids = [d.id for d in all_datasets]
    
    ids = defaultdict(list)
    challenges = defaultdict(list)
    categories = defaultdict(list)

    for _id in relevant_ids:
        ds = load_dataset(_id)["train"]
        for result, image_id in zip(ds["result"], ds["id"]):
            ids[result].append(image_id)
            challenges[parti_prompt_challenge[image_id]].append(result)
            categories[parti_prompt_categories[image_id]].append(result)
    
    all_values = sum(len(v) for v in ids.values())
    main_dict = {k: '{:.2%}'.format(len(v)/all_values) for k, v in ids.items()}
    challenges = {k: Counter(v) for k, v in challenges.items()}
    categories = {k: Counter(v) for k, v in categories.items()}

    return main_dict, challenges, categories

def sort_by_highest_percentage(df):
    # Convert percentage values to numeric format
    for column in df.columns.to_list():
        df[column] = pd.to_numeric(df[column].str.rstrip('%'))

        # Sort DataFrame by highest percentage first
        df = df.sort_values(by=column, ascending=False)

        # Convert back to percentage string format
        df[column] = df[column].astype(str) + '%'

    return df

def get_dataframe_all():
    main, challenges, categories = load_submissions()
    main_frame = pd.DataFrame([main])

    challenges_frame = pd.DataFrame.from_dict(challenges).fillna(0).T
    challenges_frame = challenges_frame.div(challenges_frame.sum(axis=1), axis=0)
    challenges_frame = challenges_frame.applymap(lambda x: '{:.2%}'.format(x))

    categories_frame = pd.DataFrame.from_dict(categories).fillna(0).T
    categories_frame = categories_frame.div(categories_frame.sum(axis=1), axis=0)
    categories_frame = categories_frame.applymap(lambda x: '{:.2%}'.format(x))

    categories_frame = categories_frame.reset_index().rename(columns={'index': 'Category'})
    challenges_frame = challenges_frame.reset_index().rename(columns={'index': 'Challenge'})

    main_frame = sort_by_highest_percentage(main_frame)

    categories_frame = categories_frame.reindex(columns=main_frame.columns.to_list())
    challenges_frame = challenges_frame.reindex(columns=main_frame.columns.to_list())

    return main_frame, challenges_frame, categories_frame

TITLE = "# Open Parti Prompts Leaderboard"
DESCRIPTION = """
*This leaderboard is retrieved from answers of [Community Evaluations on Parti Prompts](https://huggingface.co/spaces/OpenGenAI/open-parti-prompts)*
"""

EXPLANATION = """\n\n
## How the is data collected 📊 \n\n

In the [Community Parti Prompts](https://huggingface.co/spaces/OpenGenAI/open-parti-prompts), community members select for every prompt
of [Parti Prompts](https://huggingface.co/datasets/nateraw/parti-prompts) which open-source image generation model has generated the best image.
The community's answers are then stored and used in this space to give a human evaluation of the different models. \n\n

Currently the leaderboard includes the following models:
- [sd-v1-5](https://huggingface.co/runwayml/stable-diffusion-v1-5)
- [sd-v2-1](https://huggingface.co/stabilityai/stable-diffusion-2-1)
- [if-v1-0](https://huggingface.co/DeepFloyd/IF-I-XL-v1.0)
- [karlo](https://huggingface.co/kakaobrain/karlo-v1-alpha) \n\n

In the following you can see three result tables. The first shows you the overall preferences across all prompts. The second and third tables
show you a breakdown analysis per category and per type of challenge as defined by [Parti Prompts](https://huggingface.co/datasets/nateraw/parti-prompts).
"""

GALLERY_COLUMN_NUM = len(SUBMISSIONS)

def refresh():
    return get_dataframe_all()

with gr.Blocks() as demo:
    with gr.Column(visible=True) as intro_view:
        gr.Markdown(TITLE)
        gr.Markdown(DESCRIPTION)
        gr.Markdown(EXPLANATION)

    headers = list(SUBMISSIONS.keys())
    datatype = "str"

    main_df, challenge_df, category_df = get_dataframe_all()

    with gr.Column():
        gr.Markdown("# Open Parti Prompts")
        main_dataframe = gr.Dataframe(
            value=main_df,
            headers=main_df.columns.to_list(),
            datatype="str",
            row_count=main_df.shape[0],
            col_count=main_df.shape[1],
            interactive=False,
        )

    with gr.Column():
        gr.Markdown("## per category")
        cat_dataframe = gr.Dataframe(
            value=category_df,
            headers=category_df.columns.to_list(),
            datatype="str",
            row_count=category_df.shape[0],
            col_count=category_df.shape[1],
            interactive=False,
        )

    with gr.Column():
        gr.Markdown("## per challenge")
        chal_dataframe = gr.Dataframe(
            value=challenge_df,
            headers=challenge_df.columns.to_list(),
            datatype="str",
            row_count=challenge_df.shape[0],
            col_count=challenge_df.shape[1],
            interactive=False,
        )

    with gr.Row():
        refresh_button = gr.Button("Refresh")
        refresh_button.click(refresh, inputs=[], outputs=[main_dataframe, cat_dataframe, chal_dataframe])

demo.launch()