File size: 12,765 Bytes
f6a2362 c389fac f6a2362 c389fac e777562 f6a2362 c389fac f6a2362 c389fac f6a2362 c389fac f6a2362 c389fac f6a2362 c389fac f6a2362 c389fac f6a2362 c389fac f6a2362 c389fac 3951a6d f6a2362 c389fac 3951a6d c389fac 3951a6d c389fac 3951a6d c389fac 3951a6d c389fac f6a2362 3951a6d f6a2362 3951a6d c389fac f6a2362 3951a6d f6a2362 c389fac f6a2362 3951a6d f6a2362 c389fac f6a2362 3951a6d f6a2362 3951a6d f6a2362 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 |
# app.py
import json
import re
from pathlib import Path
from urllib.parse import urlparse
import streamlit as st
# ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
# Page config
# ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
st.set_page_config(
page_title="ImagenHub2 Data Visualization",
page_icon="πΌοΈ",
layout="wide",
initial_sidebar_state="expanded"
)
# ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
# Basics
# ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
MANIFEST_PATH = Path(__file__).parent / "manifest_v1.json"
DATA_DIRS = ["TIG", "TIE", "SRIG", "SRIE", "MRIG", "MRIE"]
TASK_DESCRIPTIONS = {
"TIG": "Text to Image Generation - Generate images from text prompts only",
"TIE": "Text and Image to Image Editing - Edit images based on text prompts",
"SRIG": "Single-Reference Image Generation - Generate images using text + single reference image",
"SRIE": "Single-Reference Image Editing - Edit images using text + single reference image",
"MRIG": "Multi-Reference Image Generation - Generate images using text + multiple reference images",
"MRIE": "Multi-Reference Image Editing - Edit images using text + multiple reference images"
}
DEFAULT_CHUNK = 10
if "show_counts" not in st.session_state:
st.session_state.show_counts = {task: DEFAULT_CHUNK for task in DATA_DIRS}
# ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
# Helpers: natural sorting for ids and filenames
# ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
_num_re = re.compile(r"(\d+)")
def _natural_key(s: str):
parts = _num_re.split(s)
out = []
for p in parts:
if p.isdigit():
out.append(int(p))
else:
out.append(p.lower())
return out
def _basename_from_url(url: str) -> str:
try:
return Path(urlparse(url).path).name
except Exception:
return url
def _sorted_urls(urls):
return sorted(urls, key=lambda u: _natural_key(_basename_from_url(u)))
# ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
# Load manifest (local file or baked into Space)
# ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
@st.cache_data(show_spinner=False)
def load_manifest():
with open(MANIFEST_PATH, "r", encoding="utf-8") as f:
man = json.load(f)
items = man.get("items", [])
per_task = {t: [] for t in DATA_DIRS}
topics = set()
for it in items:
# sort the image url lists inside each item for determinism
if it.get("cond_image_urls"):
it["cond_image_urls"] = _sorted_urls(it["cond_image_urls"])
if it.get("model_output_urls"):
it["model_output_urls"] = _sorted_urls(it["model_output_urls"])
per_task.setdefault(it.get("task", "Unknown"), []).append(it)
if it.get("topic"):
topics.add(it["topic"])
# sort items within each task by item_id (natural order)
for t, lst in per_task.items():
lst.sort(key=lambda it: _natural_key(str(it.get("item_id", ""))))
return per_task, sorted(list(topics))
# ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
# Stable image grid
# ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
def _display_images(urls, caption_prefix="", max_per_row=3, fixed_height_px=None):
if not urls:
st.write("No images found.")
return
cols = st.columns(max_per_row, vertical_alignment="top")
for i, url in enumerate(urls):
col = cols[i % max_per_row]
with col:
if fixed_height_px:
# Reserve space and avoid reflow while image loads
st.markdown(
f"""
<div class="img-frame" style="height:{fixed_height_px}px; display:flex; align-items:center; justify-content:center; overflow:hidden; border-radius:12px;">
<img src="{url}" alt="{_basename_from_url(url)}" style="max-height:100%; width:100%; object-fit:contain;" />
</div>
<div class="img-cap" style="font-size:0.85rem; opacity:0.8; margin-top:4px;">
{caption_prefix} {_basename_from_url(url)}
</div>
""",
unsafe_allow_html=True,
)
else:
st.image(url, caption=f"{caption_prefix} {_basename_from_url(url)}", use_container_width=True)
# ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
# Global CSS to reduce βvibratingβ / layout reflow
# ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
def _inject_css(_fixed_height_px: int | None):
css = """
<style>
.block-container { padding-top: 0.75rem; }
[data-testid="stImage"] img {
width: 100%;
height: auto;
object-fit: contain;
display: block;
}
html * {
-webkit-font-smoothing: antialiased;
-moz-osx-font-smoothing: grayscale;
text-rendering: optimizeLegibility;
}
.img-frame { background: rgba(0,0,0,0.03); }
</style>
"""
st.markdown(css, unsafe_allow_html=True)
# ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
# App
# ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
def main():
st.title("πΌοΈ ImagenHub2 Data Visualization")
st.markdown("Each task starts with **10** items β click **Show more** to load **+10**.")
# Load manifest first (to get topic list)
with st.spinner("Loading manifestβ¦"):
per_task, _all_topics = load_manifest()
# Sidebar
st.sidebar.header("Filters")
fixed_height_on = st.sidebar.toggle(
"Stabilize grid with fixed image height",
value=True,
help="Pre-allocate space for images to prevent page βvibratingβ."
)
fixed_height_px = st.sidebar.number_input(
"Fixed image height (px)",
min_value=120, max_value=1200, value=320, step=20,
disabled=not fixed_height_on
)
_inject_css(fixed_height_px if fixed_height_on else None)
selected_tasks = st.sidebar.multiselect("Select Tasks", DATA_DIRS, default=DATA_DIRS)
search_query = st.sidebar.text_input("π Search in prompts", "")
# Topic filter behaves like task filter (multiselect)
topic_filter = st.sidebar.multiselect(
"Select Topics",
_all_topics,
default=[],
help="Filter items by one or more topic IDs."
)
subtopic_filter = st.sidebar.text_input("Filter by subtopic (optional)", "")
st.sidebar.header("Task Descriptions")
for t in selected_tasks:
st.sidebar.write(f"**{t}**: {TASK_DESCRIPTIONS.get(t, '')}")
# Tabs per selected task
tabs = st.tabs(selected_tasks) if selected_tasks else []
for task, tab in zip(selected_tasks, tabs):
with tab:
st.subheader(task)
limit = st.session_state.show_counts.get(task, DEFAULT_CHUNK)
all_items = per_task.get(task, [])
# Apply filters
def _match(it):
sq = search_query.strip().lower()
if sq and (sq not in it.get("prompt", "").lower()
and sq not in it.get("prompt_refined", "").lower()):
return False
if topic_filter and it.get("topic", "") not in topic_filter:
return False
if subtopic_filter and it.get("subtopic", "") != subtopic_filter:
return False
return True
filtered = [it for it in all_items if _match(it)]
batch = filtered[:limit]
st.caption(f"Showing {len(batch)} / {len(filtered)} (from {len(all_items)} items in {task})")
if not batch:
st.warning("No items match current filters.")
else:
for it in batch:
header = f"**{it.get('item_id','?')}** β {it.get('topic','Unknown')} / {it.get('subtopic','Unknown')}"
with st.expander(header, expanded=False):
c1, c2, c3 = st.columns([1,1,1], vertical_alignment="top")
with c1: st.write(f"**Task:** {it.get('task','Unknown')}")
with c2: st.write(f"**Topic:** {it.get('topic','Unknown')}")
with c3: st.write(f"**Subtopic:** {it.get('subtopic','Unknown')}")
st.write("**Original Prompt:**")
st.write(it.get("prompt", "β"))
if it.get("prompt_refined"):
st.write("**Refined Prompt:**")
st.write(it.get("prompt_refined", ""))
if it.get("remarks"):
st.write("**Remarks:**")
st.write(it.get("remarks", ""))
cond_urls = it.get("cond_image_urls", [])
model_urls = it.get("model_output_urls", [])
if cond_urls:
st.write("**Condition Images:**")
_display_images(
cond_urls, "Condition", max_per_row=3,
fixed_height_px=(fixed_height_px if fixed_height_on else None)
)
if model_urls:
st.write("**Model Output:**")
_display_images(
model_urls, "Model", max_per_row=3,
fixed_height_px=(fixed_height_px if fixed_height_on else None)
)
st.divider()
# Pagination controls
c_more, c_reset, c_info = st.columns(3)
with c_more:
if st.button("Show more", key=f"more_{task}"):
st.session_state.show_counts[task] = limit + DEFAULT_CHUNK
st.rerun()
with c_reset:
if st.button("Reset", key=f"reset_{task}"):
st.session_state.show_counts[task] = DEFAULT_CHUNK
st.rerun()
with c_info:
st.caption(f"Current limit: {limit}")
if __name__ == "__main__":
main()
|