Spaces:
Sleeping
Sleeping
new version! multiple pages!
Browse files- app.py โ Archive/app.py +19 -10
- Archive/test.py +26 -0
- README.md +1 -1
- data/download_script.py +7 -1
- pages/1_๐ผ๏ธ_Gallery.py +348 -0
- pages/2_๐๏ธ_Ranking.py +28 -0
- ๐ _Home.py +48 -0
app.py โ Archive/app.py
RENAMED
|
@@ -100,10 +100,10 @@ class GalleryApp:
|
|
| 100 |
|
| 101 |
st.image(image, use_column_width=True)
|
| 102 |
|
| 103 |
-
#
|
| 104 |
-
|
| 105 |
-
|
| 106 |
-
|
| 107 |
|
| 108 |
# st.write(idx+j)
|
| 109 |
# show selected info
|
|
@@ -294,8 +294,8 @@ class GalleryApp:
|
|
| 294 |
return items, info, col_num
|
| 295 |
|
| 296 |
def app(self):
|
| 297 |
-
st.title('Model
|
| 298 |
-
st.write('This is a gallery of images generated by the models
|
| 299 |
|
| 300 |
with st.sidebar:
|
| 301 |
prompt_tags = self.promptBook['tag'].unique()
|
|
@@ -367,7 +367,7 @@ class GalleryApp:
|
|
| 367 |
|
| 368 |
with st.form(key=f'{prompt_id}', clear_on_submit=True):
|
| 369 |
# buttons = st.columns([1, 1, 1])
|
| 370 |
-
buttons_space = st.
|
| 371 |
gallery_space = st.empty()
|
| 372 |
# with buttons[0]:
|
| 373 |
# submit = st.form_submit_button('Save selections', on_click=self.save_checked, use_container_width=True, type='primary')
|
|
@@ -379,8 +379,17 @@ class GalleryApp:
|
|
| 379 |
with gallery_space.container():
|
| 380 |
self.gallery_standard(items, col_num, info)
|
| 381 |
|
| 382 |
-
with buttons_space:
|
| 383 |
-
st.form_submit_button('
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 384 |
|
| 385 |
|
| 386 |
def reset_current_prompt(self, prompt_id):
|
|
@@ -416,7 +425,7 @@ def load_hf_dataset():
|
|
| 416 |
# load from huggingface
|
| 417 |
roster = pd.DataFrame(load_dataset('NYUSHPRP/ModelCofferRoster', split='train'))
|
| 418 |
promptBook = pd.DataFrame(load_dataset('NYUSHPRP/ModelCofferMetadata', split='train'))
|
| 419 |
-
images_ds = load_from_disk(os.path.join(os.getcwd(), 'data', 'promptbook'))
|
| 420 |
|
| 421 |
# process dataset
|
| 422 |
roster = roster[['model_id', 'model_name', 'modelVersion_id', 'modelVersion_name',
|
|
|
|
| 100 |
|
| 101 |
st.image(image, use_column_width=True)
|
| 102 |
|
| 103 |
+
# show checkbox
|
| 104 |
+
self.promptBook.loc[items.iloc[idx + j]['row_idx'].item(), 'checked'] = st.checkbox(
|
| 105 |
+
'Select', value=self.promptBook.loc[items.iloc[idx + j]['row_idx'].item(), 'checked'],
|
| 106 |
+
key=f'select_{idx + j}')
|
| 107 |
|
| 108 |
# st.write(idx+j)
|
| 109 |
# show selected info
|
|
|
|
| 294 |
return items, info, col_num
|
| 295 |
|
| 296 |
def app(self):
|
| 297 |
+
st.title('Model Visualization and Retrieval')
|
| 298 |
+
st.write('This is a gallery of images generated by the models')
|
| 299 |
|
| 300 |
with st.sidebar:
|
| 301 |
prompt_tags = self.promptBook['tag'].unique()
|
|
|
|
| 367 |
|
| 368 |
with st.form(key=f'{prompt_id}', clear_on_submit=True):
|
| 369 |
# buttons = st.columns([1, 1, 1])
|
| 370 |
+
buttons_space = st.columns([1, 1, 1, 1])
|
| 371 |
gallery_space = st.empty()
|
| 372 |
# with buttons[0]:
|
| 373 |
# submit = st.form_submit_button('Save selections', on_click=self.save_checked, use_container_width=True, type='primary')
|
|
|
|
| 379 |
with gallery_space.container():
|
| 380 |
self.gallery_standard(items, col_num, info)
|
| 381 |
|
| 382 |
+
with buttons_space[0]:
|
| 383 |
+
st.form_submit_button('Confirm and Continue', use_container_width=True, type='primary')
|
| 384 |
+
|
| 385 |
+
with buttons_space[1]:
|
| 386 |
+
st.form_submit_button('Select All', use_container_width=True)
|
| 387 |
+
|
| 388 |
+
with buttons_space[2]:
|
| 389 |
+
st.form_submit_button('Deselect All', use_container_width=True)
|
| 390 |
+
|
| 391 |
+
with buttons_space[3]:
|
| 392 |
+
st.form_submit_button('Refresh', on_click=gallery_space.empty, use_container_width=True)
|
| 393 |
|
| 394 |
|
| 395 |
def reset_current_prompt(self, prompt_id):
|
|
|
|
| 425 |
# load from huggingface
|
| 426 |
roster = pd.DataFrame(load_dataset('NYUSHPRP/ModelCofferRoster', split='train'))
|
| 427 |
promptBook = pd.DataFrame(load_dataset('NYUSHPRP/ModelCofferMetadata', split='train'))
|
| 428 |
+
images_ds = load_from_disk(os.path.join(os.getcwd(), '../data', 'promptbook'))
|
| 429 |
|
| 430 |
# process dataset
|
| 431 |
roster = roster[['model_id', 'model_name', 'modelVersion_id', 'modelVersion_name',
|
Archive/test.py
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import streamlit as st
|
| 2 |
+
|
| 3 |
+
if __name__ == "__main__":
|
| 4 |
+
if 'check_dict' not in st.session_state:
|
| 5 |
+
st.session_state.check_dict = {'check1': False, 'check2': False, 'check3': False}
|
| 6 |
+
|
| 7 |
+
with st.form('my_form'):
|
| 8 |
+
st.session_state.check_dict['check1'] = st.checkbox('Check 1 out')
|
| 9 |
+
st.session_state.check_dict['check2'] = st.checkbox('Check 2 out')
|
| 10 |
+
st.session_state.check_dict['check3'] = st.checkbox('Check 3 out')
|
| 11 |
+
|
| 12 |
+
check21 = st.checkbox('Check 21 out')
|
| 13 |
+
if check21:
|
| 14 |
+
st.write('check21 is checked')
|
| 15 |
+
check22 = st.checkbox('Check 22 out')
|
| 16 |
+
if check22:
|
| 17 |
+
st.write('check22 is checked')
|
| 18 |
+
check23 = st.checkbox('Check 23 out')
|
| 19 |
+
if check23:
|
| 20 |
+
st.write('check23 is checked')
|
| 21 |
+
|
| 22 |
+
# Every form must have a submit button.
|
| 23 |
+
submitted = st.form_submit_button('Submit')
|
| 24 |
+
|
| 25 |
+
for key, value in st.session_state.check_dict.items():
|
| 26 |
+
st.write(key, value)
|
README.md
CHANGED
|
@@ -6,7 +6,7 @@ colorTo: purple
|
|
| 6 |
sdk: streamlit
|
| 7 |
sdk_version: 1.19.0
|
| 8 |
python_version: 3.9.13
|
| 9 |
-
app_file:
|
| 10 |
pinned: false
|
| 11 |
---
|
| 12 |
|
|
|
|
| 6 |
sdk: streamlit
|
| 7 |
sdk_version: 1.19.0
|
| 8 |
python_version: 3.9.13
|
| 9 |
+
app_file: ๐ _Home.py
|
| 10 |
pinned: false
|
| 11 |
---
|
| 12 |
|
data/download_script.py
CHANGED
|
@@ -20,5 +20,11 @@ def test():
|
|
| 20 |
print(promptbook[0]['image'])
|
| 21 |
|
| 22 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 23 |
if __name__ == '__main__':
|
| 24 |
-
main()
|
|
|
|
| 20 |
print(promptbook[0]['image'])
|
| 21 |
|
| 22 |
|
| 23 |
+
# def drop_metadata_checked_column():
|
| 24 |
+
# ModelCofferMetadata = load_dataset('NYUSHPRP/ModelCofferMetadata', split='train')
|
| 25 |
+
# ModelCofferMetadata = ModelCofferMetadata.remove_columns(['checked'])
|
| 26 |
+
# ModelCofferMetadata.push_to_hub('NYUSHPRP/ModelCofferMetadata', split='train')
|
| 27 |
+
|
| 28 |
+
|
| 29 |
if __name__ == '__main__':
|
| 30 |
+
main()
|
pages/1_๐ผ๏ธ_Gallery.py
ADDED
|
@@ -0,0 +1,348 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import streamlit as st
|
| 2 |
+
import numpy as np
|
| 3 |
+
import pandas as pd
|
| 4 |
+
import glob
|
| 5 |
+
from datasets import load_dataset, Dataset, load_from_disk
|
| 6 |
+
from huggingface_hub import login
|
| 7 |
+
import os
|
| 8 |
+
import requests
|
| 9 |
+
from bs4 import BeautifulSoup
|
| 10 |
+
import altair as alt
|
| 11 |
+
from streamlit_extras.switch_page_button import switch_page
|
| 12 |
+
|
| 13 |
+
SCORE_NAME_MAPPING = {'clip': 'clip_score', 'rank': 'avg_rank', 'pop': 'model_download_count'}
|
| 14 |
+
|
| 15 |
+
|
| 16 |
+
# hist_data = pd.DataFrame(np.random.normal(42, 10, (200, 1)), columns=["x"])
|
| 17 |
+
@st.cache_resource
|
| 18 |
+
def altair_histogram(hist_data, sort_by, mini, maxi):
|
| 19 |
+
brushed = alt.selection_interval(encodings=['x'], name="brushed")
|
| 20 |
+
|
| 21 |
+
chart = (
|
| 22 |
+
alt.Chart(hist_data)
|
| 23 |
+
.mark_bar(opacity=0.7, cornerRadius=2)
|
| 24 |
+
.encode(alt.X(f"{sort_by}:Q", bin=alt.Bin(maxbins=25)), y="count()")
|
| 25 |
+
# .add_selection(brushed)
|
| 26 |
+
# .properties(width=800, height=300)
|
| 27 |
+
)
|
| 28 |
+
|
| 29 |
+
# Create a transparent rectangle for highlighting the range
|
| 30 |
+
highlight = (
|
| 31 |
+
alt.Chart(pd.DataFrame({'x1': [mini], 'x2': [maxi]}))
|
| 32 |
+
.mark_rect(opacity=0.3)
|
| 33 |
+
.encode(x='x1', x2='x2')
|
| 34 |
+
# .properties(width=800, height=300)
|
| 35 |
+
)
|
| 36 |
+
|
| 37 |
+
# Layer the chart and the highlight rectangle
|
| 38 |
+
layered_chart = alt.layer(chart, highlight)
|
| 39 |
+
|
| 40 |
+
return layered_chart
|
| 41 |
+
|
| 42 |
+
|
| 43 |
+
class GalleryApp:
|
| 44 |
+
def __init__(self, promptBook, images_ds):
|
| 45 |
+
self.promptBook = promptBook
|
| 46 |
+
self.images_ds = images_ds
|
| 47 |
+
|
| 48 |
+
def gallery_standard(self, items, col_num, info):
|
| 49 |
+
rows = len(items) // col_num + 1
|
| 50 |
+
containers = [st.container() for _ in range(rows)]
|
| 51 |
+
for idx in range(0, len(items), col_num):
|
| 52 |
+
row_idx = idx // col_num
|
| 53 |
+
with containers[row_idx]:
|
| 54 |
+
cols = st.columns(col_num)
|
| 55 |
+
for j in range(col_num):
|
| 56 |
+
if idx + j < len(items):
|
| 57 |
+
with cols[j]:
|
| 58 |
+
# show image
|
| 59 |
+
image = self.images_ds[items.iloc[idx + j]['row_idx'].item()]['image']
|
| 60 |
+
st.image(image, use_column_width=True)
|
| 61 |
+
|
| 62 |
+
# handel checkbox information
|
| 63 |
+
prompt_id = items.iloc[idx + j]['prompt_id']
|
| 64 |
+
modelVersion_id = items.iloc[idx + j]['modelVersion_id']
|
| 65 |
+
check_init = True if modelVersion_id in st.session_state.selected_dict.get(prompt_id, []) else False
|
| 66 |
+
|
| 67 |
+
# show checkbox
|
| 68 |
+
checked = st.checkbox('Select', key=f'select_{idx + j}', value=check_init)
|
| 69 |
+
if checked:
|
| 70 |
+
st.session_state.selected_dict[prompt_id] = st.session_state.selected_dict.get(prompt_id, []) + [modelVersion_id]
|
| 71 |
+
else:
|
| 72 |
+
try:
|
| 73 |
+
st.session_state.selected_dict[prompt_id].remove(modelVersion_id)
|
| 74 |
+
except:
|
| 75 |
+
pass
|
| 76 |
+
|
| 77 |
+
# show selected info
|
| 78 |
+
for key in info:
|
| 79 |
+
st.write(f"**{key}**: {items.iloc[idx + j][key]}")
|
| 80 |
+
|
| 81 |
+
def selection_panel(self, items):
|
| 82 |
+
selecters = st.columns([1, 4])
|
| 83 |
+
|
| 84 |
+
# select sort type
|
| 85 |
+
with selecters[0]:
|
| 86 |
+
sort_type = st.selectbox('Sort by', ['Scores', 'IDs and Names'])
|
| 87 |
+
if sort_type == 'Scores':
|
| 88 |
+
sort_by = 'weighted_score_sum'
|
| 89 |
+
|
| 90 |
+
# select other options
|
| 91 |
+
with selecters[1]:
|
| 92 |
+
if sort_type == 'IDs and Names':
|
| 93 |
+
sub_selecters = st.columns([3, 1])
|
| 94 |
+
# select sort by
|
| 95 |
+
with sub_selecters[0]:
|
| 96 |
+
sort_by = st.selectbox('Sort by',
|
| 97 |
+
['model_name', 'model_id', 'modelVersion_name', 'modelVersion_id'],
|
| 98 |
+
label_visibility='hidden')
|
| 99 |
+
|
| 100 |
+
continue_idx = 1
|
| 101 |
+
|
| 102 |
+
else:
|
| 103 |
+
# add custom weights
|
| 104 |
+
sub_selecters = st.columns([1, 1, 1, 1])
|
| 105 |
+
|
| 106 |
+
if 'score_weights' not in st.session_state:
|
| 107 |
+
st.session_state.score_weights = [1.0, 0.8, 0.2, 0.84]
|
| 108 |
+
|
| 109 |
+
with sub_selecters[0]:
|
| 110 |
+
clip_weight = st.number_input('Clip Score Weight', min_value=-100.0, max_value=100.0, value=st.session_state.score_weights[0], step=0.1, help='the weight for normalized clip score')
|
| 111 |
+
with sub_selecters[1]:
|
| 112 |
+
rank_weight = st.number_input('Distinctiveness Weight', min_value=-100.0, max_value=100.0, value=st.session_state.score_weights[1], step=0.1, help='the weight for average rank')
|
| 113 |
+
with sub_selecters[2]:
|
| 114 |
+
pop_weight = st.number_input('Popularity Weight', min_value=-100.0, max_value=100.0, value=st.session_state.score_weights[2], step=0.1, help='the weight for normalized popularity score')
|
| 115 |
+
|
| 116 |
+
items.loc[:, 'weighted_score_sum'] = round(items['norm_clip'] * clip_weight + items['avg_rank'] * rank_weight + items[
|
| 117 |
+
'norm_pop'] * pop_weight, 4)
|
| 118 |
+
|
| 119 |
+
continue_idx = 3
|
| 120 |
+
|
| 121 |
+
# select threshold
|
| 122 |
+
with sub_selecters[continue_idx]:
|
| 123 |
+
dist_threshold = st.number_input('Distinctiveness Threshold', min_value=0.0, max_value=1.0, value=st.session_state.score_weights[3], step=0.01, help='Only show models with distinctiveness score lower than this threshold, set 1.0 to show all images')
|
| 124 |
+
items = items[items['avg_rank'] < dist_threshold].reset_index(drop=True)
|
| 125 |
+
|
| 126 |
+
# save latest weights
|
| 127 |
+
st.session_state.score_weights = [clip_weight, rank_weight, pop_weight, dist_threshold]
|
| 128 |
+
|
| 129 |
+
# draw a distribution histogram
|
| 130 |
+
if sort_type == 'Scores':
|
| 131 |
+
try:
|
| 132 |
+
with st.expander('Show score distribution histogram and select score range'):
|
| 133 |
+
st.write('**Score distribution histogram**')
|
| 134 |
+
chart_space = st.container()
|
| 135 |
+
# st.write('Select the range of scores to show')
|
| 136 |
+
hist_data = pd.DataFrame(items[sort_by])
|
| 137 |
+
mini = hist_data[sort_by].min().item()
|
| 138 |
+
mini = mini//0.1 * 0.1
|
| 139 |
+
maxi = hist_data[sort_by].max().item()
|
| 140 |
+
maxi = maxi//0.1 * 0.1 + 0.1
|
| 141 |
+
st.write('**Select the range of scores to show**')
|
| 142 |
+
r = st.slider('Select the range of scores to show', min_value=mini, max_value=maxi, value=(mini, maxi), step=0.05, label_visibility='collapsed')
|
| 143 |
+
with chart_space:
|
| 144 |
+
st.altair_chart(altair_histogram(hist_data, sort_by, r[0], r[1]), use_container_width=True)
|
| 145 |
+
# event_dict = altair_component(altair_chart=altair_histogram(hist_data, sort_by))
|
| 146 |
+
# r = event_dict.get(sort_by)
|
| 147 |
+
if r:
|
| 148 |
+
items = items[(items[sort_by] >= r[0]) & (items[sort_by] <= r[1])].reset_index(drop=True)
|
| 149 |
+
# st.write(r)
|
| 150 |
+
except:
|
| 151 |
+
pass
|
| 152 |
+
|
| 153 |
+
display_options = st.columns([1, 4])
|
| 154 |
+
|
| 155 |
+
with display_options[0]:
|
| 156 |
+
# select order
|
| 157 |
+
order = st.selectbox('Order', ['Ascending', 'Descending'], index=1 if sort_type == 'Scores' else 0)
|
| 158 |
+
if order == 'Ascending':
|
| 159 |
+
order = True
|
| 160 |
+
else:
|
| 161 |
+
order = False
|
| 162 |
+
|
| 163 |
+
with display_options[1]:
|
| 164 |
+
|
| 165 |
+
# select info to show
|
| 166 |
+
info = st.multiselect('Show Info',
|
| 167 |
+
['model_download_count', 'clip_score', 'avg_rank', 'model_name', 'model_id',
|
| 168 |
+
'modelVersion_name', 'modelVersion_id', 'clip+rank', 'clip+pop', 'rank+pop',
|
| 169 |
+
'clip+rank+pop', 'weighted_score_sum'],
|
| 170 |
+
default=sort_by)
|
| 171 |
+
|
| 172 |
+
# apply sorting to dataframe
|
| 173 |
+
items = items.sort_values(by=[sort_by], ascending=order).reset_index(drop=True)
|
| 174 |
+
|
| 175 |
+
# select number of columns
|
| 176 |
+
col_num = st.slider('Number of columns', min_value=1, max_value=9, value=4, step=1, key='col_num')
|
| 177 |
+
|
| 178 |
+
return items, info, col_num
|
| 179 |
+
|
| 180 |
+
def sidebar(self):
|
| 181 |
+
with st.sidebar:
|
| 182 |
+
prompt_tags = self.promptBook['tag'].unique()
|
| 183 |
+
# sort tags by alphabetical order
|
| 184 |
+
prompt_tags = np.sort(prompt_tags)[::-1]
|
| 185 |
+
|
| 186 |
+
tag = st.selectbox('Select a tag', prompt_tags)
|
| 187 |
+
|
| 188 |
+
items = self.promptBook[self.promptBook['tag'] == tag].reset_index(drop=True)
|
| 189 |
+
|
| 190 |
+
original_prompts = np.sort(items['prompt'].unique())[::-1]
|
| 191 |
+
|
| 192 |
+
# remove the first four items in the prompt, which are mostly the same
|
| 193 |
+
if tag != 'abstract':
|
| 194 |
+
prompts = [', '.join(x.split(', ')[4:]) for x in original_prompts]
|
| 195 |
+
prompt = st.selectbox('Select prompt', prompts)
|
| 196 |
+
|
| 197 |
+
idx = prompts.index(prompt)
|
| 198 |
+
prompt_full = ', '.join(original_prompts[idx].split(', ')[:4]) + ', ' + prompt
|
| 199 |
+
else:
|
| 200 |
+
prompt_full = st.selectbox('Select prompt', original_prompts)
|
| 201 |
+
|
| 202 |
+
items = items[items['prompt'] == prompt_full].reset_index(drop=True)
|
| 203 |
+
prompt_id = items['prompt_id'].unique()[0]
|
| 204 |
+
|
| 205 |
+
# show image metadata
|
| 206 |
+
image_metadatas = ['prompt_id', 'prompt', 'negativePrompt', 'sampler', 'cfgScale', 'size', 'seed']
|
| 207 |
+
for key in image_metadatas:
|
| 208 |
+
label = ' '.join(key.split('_')).capitalize()
|
| 209 |
+
st.write(f"**{label}**")
|
| 210 |
+
if items[key][0] == ' ':
|
| 211 |
+
st.write('`None`')
|
| 212 |
+
else:
|
| 213 |
+
st.caption(f"{items[key][0]}")
|
| 214 |
+
|
| 215 |
+
# for tag as civitai, add civitai reference
|
| 216 |
+
if tag == 'civitai':
|
| 217 |
+
try:
|
| 218 |
+
st.write('**Civitai Reference**')
|
| 219 |
+
res = requests.get(f'https://civitai.com/images/{prompt_id.item()}')
|
| 220 |
+
# st.write(res.text)
|
| 221 |
+
soup = BeautifulSoup(res.text, 'html.parser')
|
| 222 |
+
image_section = soup.find('div', {'class': 'mantine-12rlksp'})
|
| 223 |
+
image_url = image_section.find('img')['src']
|
| 224 |
+
st.image(image_url, use_column_width=True)
|
| 225 |
+
except:
|
| 226 |
+
pass
|
| 227 |
+
|
| 228 |
+
return prompt_tags, tag, prompt_id, items
|
| 229 |
+
|
| 230 |
+
def app(self):
|
| 231 |
+
st.title('Model Visualization and Retrieval')
|
| 232 |
+
st.write('This is a gallery of images generated by the models')
|
| 233 |
+
|
| 234 |
+
prompt_tags, tag, prompt_id, items = self.sidebar()
|
| 235 |
+
|
| 236 |
+
# add safety check for some prompts
|
| 237 |
+
safety_check = True
|
| 238 |
+
unsafe_prompts = {}
|
| 239 |
+
# initialize unsafe prompts
|
| 240 |
+
for prompt_tag in prompt_tags:
|
| 241 |
+
unsafe_prompts[prompt_tag] = []
|
| 242 |
+
# manually add unsafe prompts
|
| 243 |
+
unsafe_prompts['civitai'] = [375790, 366222, 295008, 256477]
|
| 244 |
+
unsafe_prompts['people'] = [53]
|
| 245 |
+
unsafe_prompts['art'] = [23]
|
| 246 |
+
unsafe_prompts['abstract'] = [10, 12]
|
| 247 |
+
unsafe_prompts['food'] = [34]
|
| 248 |
+
|
| 249 |
+
if int(prompt_id.item()) in unsafe_prompts[tag]:
|
| 250 |
+
st.warning('This prompt may contain unsafe content. They might be offensive, depressing, or sexual.')
|
| 251 |
+
safety_check = st.checkbox('I understand that this prompt may contain unsafe content. Show these images anyway.', key=f'{prompt_id}')
|
| 252 |
+
|
| 253 |
+
if safety_check:
|
| 254 |
+
items, info, col_num = self.selection_panel(items)
|
| 255 |
+
# self.gallery_standard(items, col_num, info)
|
| 256 |
+
|
| 257 |
+
with st.form(key=f'{prompt_id}'):
|
| 258 |
+
# buttons = st.columns([1, 1, 1])
|
| 259 |
+
buttons_space = st.columns([1, 1, 1, 1])
|
| 260 |
+
gallery_space = st.empty()
|
| 261 |
+
|
| 262 |
+
with buttons_space[0]:
|
| 263 |
+
continue_btn = st.form_submit_button('Confirm Selection', use_container_width=True, type='primary')
|
| 264 |
+
if continue_btn:
|
| 265 |
+
self.submit_actions('Continue', prompt_id)
|
| 266 |
+
|
| 267 |
+
with buttons_space[1]:
|
| 268 |
+
select_btn = st.form_submit_button('Select All', use_container_width=True)
|
| 269 |
+
if select_btn:
|
| 270 |
+
self.submit_actions('Select', prompt_id)
|
| 271 |
+
|
| 272 |
+
with buttons_space[2]:
|
| 273 |
+
deselect_btn = st.form_submit_button('Deselect All', use_container_width=True)
|
| 274 |
+
if deselect_btn:
|
| 275 |
+
self.submit_actions('Deselect', prompt_id)
|
| 276 |
+
|
| 277 |
+
with buttons_space[3]:
|
| 278 |
+
refresh_btn = st.form_submit_button('Refresh', on_click=gallery_space.empty, use_container_width=True)
|
| 279 |
+
|
| 280 |
+
with gallery_space.container():
|
| 281 |
+
with st.spinner('Loading images...'):
|
| 282 |
+
self.gallery_standard(items, col_num, info)
|
| 283 |
+
|
| 284 |
+
def submit_actions(self, status, prompt_id):
|
| 285 |
+
if status == 'Select':
|
| 286 |
+
modelVersions = self.promptBook[self.promptBook['prompt_id'] == prompt_id]['modelVersion_id'].unique()
|
| 287 |
+
st.session_state.selected_dict[prompt_id] = modelVersions.tolist()
|
| 288 |
+
print(st.session_state.selected_dict, 'select')
|
| 289 |
+
elif status == 'Deselect':
|
| 290 |
+
st.session_state.selected_dict[prompt_id] = []
|
| 291 |
+
print(st.session_state.selected_dict, 'deselect')
|
| 292 |
+
# self.promptBook.loc[self.promptBook['prompt_id'] == prompt_id, 'checked'] = False
|
| 293 |
+
pass
|
| 294 |
+
elif status == 'Continue':
|
| 295 |
+
# switch_page("ranking")
|
| 296 |
+
pass
|
| 297 |
+
|
| 298 |
+
|
| 299 |
+
@st.cache_data
|
| 300 |
+
def load_hf_dataset():
|
| 301 |
+
# login to huggingface
|
| 302 |
+
login(token=os.environ.get("HF_TOKEN"))
|
| 303 |
+
|
| 304 |
+
# load from huggingface
|
| 305 |
+
roster = pd.DataFrame(load_dataset('NYUSHPRP/ModelCofferRoster', split='train'))
|
| 306 |
+
promptBook = pd.DataFrame(load_dataset('NYUSHPRP/ModelCofferMetadata', split='train'))
|
| 307 |
+
images_ds = load_from_disk(os.path.join(os.getcwd(), 'data', 'promptbook'))
|
| 308 |
+
|
| 309 |
+
# process dataset
|
| 310 |
+
roster = roster[['model_id', 'model_name', 'modelVersion_id', 'modelVersion_name',
|
| 311 |
+
'model_download_count']].drop_duplicates().reset_index(drop=True)
|
| 312 |
+
|
| 313 |
+
# # add 'checked' column to promptBook if not exist
|
| 314 |
+
# if 'checked' not in promptBook.columns:
|
| 315 |
+
# promptBook.loc[:, 'checked'] = False
|
| 316 |
+
|
| 317 |
+
# add 'custom_score_weights' column to promptBook if not exist
|
| 318 |
+
if 'weighted_score_sum' not in promptBook.columns:
|
| 319 |
+
promptBook.loc[:, 'weighted_score_sum'] = 0
|
| 320 |
+
|
| 321 |
+
# merge roster and promptbook
|
| 322 |
+
promptBook = promptBook.merge(roster[['model_id', 'model_name', 'modelVersion_id', 'modelVersion_name', 'model_download_count']],
|
| 323 |
+
on=['model_id', 'modelVersion_id'], how='left')
|
| 324 |
+
|
| 325 |
+
# add column to record current row index
|
| 326 |
+
promptBook.loc[:, 'row_idx'] = promptBook.index
|
| 327 |
+
|
| 328 |
+
return roster, promptBook, images_ds
|
| 329 |
+
|
| 330 |
+
|
| 331 |
+
if __name__ == "__main__":
|
| 332 |
+
st.set_page_config(page_title="Model Coffer Gallery", page_icon="๐ผ๏ธ", layout="wide")
|
| 333 |
+
if 'user_id' not in st.session_state:
|
| 334 |
+
st.warning('Please log in first.')
|
| 335 |
+
home_btn = st.button('Go to Home Page')
|
| 336 |
+
if home_btn:
|
| 337 |
+
switch_page("home")
|
| 338 |
+
else:
|
| 339 |
+
st.write('You have already logged in as ' + st.session_state.user_id[0])
|
| 340 |
+
roster, promptBook, st.session_state["images_ds"] = load_hf_dataset()
|
| 341 |
+
# print(promptBook.columns)
|
| 342 |
+
|
| 343 |
+
# initialize selected_dict
|
| 344 |
+
if 'selected_dict' not in st.session_state:
|
| 345 |
+
st.session_state['selected_dict'] = {}
|
| 346 |
+
|
| 347 |
+
app = GalleryApp(promptBook=promptBook, images_ds=st.session_state.images_ds)
|
| 348 |
+
app.app()
|
pages/2_๐๏ธ_Ranking.py
ADDED
|
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import streamlit as st
|
| 2 |
+
import numpy as np
|
| 3 |
+
import pandas as pd
|
| 4 |
+
from streamlit_extras.switch_page_button import switch_page
|
| 5 |
+
|
| 6 |
+
if __name__ == "__main__":
|
| 7 |
+
st.set_page_config(page_title="Personal Image Ranking", page_icon="๐๏ธ๏ธ", layout="wide")
|
| 8 |
+
|
| 9 |
+
if 'user_id' not in st.session_state:
|
| 10 |
+
st.warning('Please log in first.')
|
| 11 |
+
home_btn = st.button('Go to Home Page')
|
| 12 |
+
if home_btn:
|
| 13 |
+
switch_page("home")
|
| 14 |
+
|
| 15 |
+
else:
|
| 16 |
+
all_checked = 0
|
| 17 |
+
for key, value in st.session_state.selected_dict.items():
|
| 18 |
+
for v in value:
|
| 19 |
+
all_checked += 1
|
| 20 |
+
|
| 21 |
+
if all_checked == 0:
|
| 22 |
+
st.info('You have not checked any image yet. Please go back to the gallery page and check some images.')
|
| 23 |
+
gallery_btn = st.button('Go to Gallery')
|
| 24 |
+
if gallery_btn:
|
| 25 |
+
switch_page('gallery')
|
| 26 |
+
else:
|
| 27 |
+
st.write('You have checked ' + str(all_checked) + ' images.')
|
| 28 |
+
|
๐ _Home.py
ADDED
|
@@ -0,0 +1,48 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import streamlit as st
|
| 2 |
+
import random
|
| 3 |
+
import time
|
| 4 |
+
from streamlit_extras.switch_page_button import switch_page
|
| 5 |
+
|
| 6 |
+
|
| 7 |
+
def login():
|
| 8 |
+
# skip customize user name for debug mode
|
| 9 |
+
|
| 10 |
+
with st.form("user_login"):
|
| 11 |
+
st.write('## Enter Your Name')
|
| 12 |
+
user_id = st.text_input(
|
| 13 |
+
"Enter your name for personalization ๐",
|
| 14 |
+
label_visibility='visible',
|
| 15 |
+
disabled=False,
|
| 16 |
+
placeholder='anonymous',
|
| 17 |
+
)
|
| 18 |
+
st.write('You can leave it blank to be anonymous.')
|
| 19 |
+
|
| 20 |
+
# Every form must have a submit button.
|
| 21 |
+
submitted = st.form_submit_button("Start")
|
| 22 |
+
if submitted:
|
| 23 |
+
save_user_id(user_id)
|
| 24 |
+
switch_page("gallery")
|
| 25 |
+
|
| 26 |
+
|
| 27 |
+
def save_user_id(user_id):
|
| 28 |
+
print(user_id)
|
| 29 |
+
if not user_id:
|
| 30 |
+
user_id = 'anonymous' + str(random.randint(0, 100000))
|
| 31 |
+
st.session_state.user_id = [user_id, time.time()]
|
| 32 |
+
|
| 33 |
+
|
| 34 |
+
if __name__ == '__main__':
|
| 35 |
+
st.set_page_config(page_title="Login", page_icon="๐ ")
|
| 36 |
+
|
| 37 |
+
st.title("Personalized Image Ranking")
|
| 38 |
+
st.write(
|
| 39 |
+
"This is an web application to collect personal preference to ai generated images. \
|
| 40 |
+
You can know which model you like most after you finish the survey."
|
| 41 |
+
)
|
| 42 |
+
|
| 43 |
+
if 'user_id' not in st.session_state:
|
| 44 |
+
login()
|
| 45 |
+
else:
|
| 46 |
+
st.write('You have already logged in as ' + st.session_state.user_id[0])
|
| 47 |
+
st.button('Log out', on_click=lambda: st.session_state.pop('user_id'))
|
| 48 |
+
|