File size: 21,120 Bytes
6371d28 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 | """
Dataset Visualizer
MMB-style CSVs from upload, output/, or data/.
Image sets table includes thumbnails; overview, difficulty & questions, answer matrix.
"""
from __future__ import annotations
import io
import json
import shutil
import tempfile
import zipfile
import streamlit as st
import pandas as pd
import plotly.express as px
from pathlib import Path
st.set_page_config(page_title="Dataset Visualizer", layout="wide")
PROJECT_ROOT = Path(__file__).resolve().parent
OUTPUT_DIR = PROJECT_ROOT / "output"
DATA_DIR = PROJECT_ROOT / "data"
HF_DATASET_DIR = PROJECT_ROOT / "hf_dataset" # For Hugging Face Space / repo bundle
WORKSPACE_ROOT = PROJECT_ROOT
# MMB CSV markers
MMB_IMAGE_COLS = ["original_image", "counterfactual1_image", "counterfactual2_image"]
MMB_QA_COLS = ["original_question", "counterfactual1_question", "counterfactual2_question"]
MMB_DIFF_COLS = [
"original_question_difficulty",
"counterfactual1_question_difficulty",
"counterfactual2_question_difficulty",
]
def resolve_image_path(csv_path: Path, fname: str) -> Path | None:
"""Try multiple locations; return first existing path for fname."""
base = csv_path.resolve().parent
candidates = [
base / "images" / fname,
base / fname,
base.parent / "images" / fname,
base.parent / fname,
]
for p in candidates:
if p.exists() and p.is_file():
return p
return None
def scene_id_from_original(original_image: str) -> str:
"""e.g. 'scene_0001_original.png' -> 'scene_0001'."""
s = str(original_image)
for suf in ("_original.png", "_original", ".png"):
if s.endswith(suf):
s = s[: -len(suf)]
break
return s or original_image
def scene_id_from_image_name(fname: str) -> str:
"""e.g. 'scene_0001_cf1.png' or 'scene_0001_original.png' -> 'scene_0001'. Handles _original, _cf1, _cf2."""
s = str(fname).strip()
for suf in (
"_original.png", "_original",
"_cf1.png", "_cf1",
"_cf2.png", "_cf2",
".png",
):
if s.lower().endswith(suf.lower()):
s = s[: -len(suf)]
break
return s.strip() or fname
def _row_scene_id(row, index: int) -> str:
"""Scene ID for display from first image column."""
v = row.get(MMB_IMAGE_COLS[0], None)
if isinstance(v, str) and v:
return scene_id_from_original(v)
return f"row_{index + 1}"
def resolve_scenes_dir(csv_path: Path) -> Path | None:
"""Scenes dir: <csv_parent>/scenes/ or <csv_grandparent>/scenes/."""
base = csv_path.resolve().parent
for candidate in (base / "scenes", base.parent / "scenes"):
if candidate.is_dir():
return candidate
return None
def _scene_id_for_lookup(s: str) -> str:
"""Normalize scene id for scene file lookup (lowercase, strip)."""
return str(s).strip().lower() if s else ""
@st.cache_data
def _get_cf_type_from_scene_file(csv_path: Path, scene_id: str, variant: str) -> str | None:
"""Load scenes/{scene_id}_{variant}.json; return cf_metadata.cf_type. Uses variant from JSON to validate."""
scenes_dir = resolve_scenes_dir(csv_path)
if not scenes_dir:
return None
sid = _scene_id_for_lookup(scene_id)
if not sid:
return None
path = scenes_dir / f"{sid}_{variant}.json"
if not path.exists():
return None
try:
with open(path, encoding="utf-8") as f:
data = json.load(f)
meta = data.get("cf_metadata") or {}
v = meta.get("variant")
if v is not None and str(v).lower() != variant.lower():
return None
t = meta.get("cf_type")
return str(t) if t is not None else None
except Exception:
return None
def get_cf_types_for_scene(csv_path: Path | None, row) -> tuple[str | None, str | None]:
"""Load scene JSONs for this row; return (cf1_type, cf2_type). Uses CF image columns to resolve scene."""
if csv_path is None:
return (None, None)
sid_orig = scene_id_from_original(str(row.get(MMB_IMAGE_COLS[0], "") or ""))
sid_cf1 = scene_id_from_image_name(str(row.get(MMB_IMAGE_COLS[1], "") or "")) or sid_orig
sid_cf2 = scene_id_from_image_name(str(row.get(MMB_IMAGE_COLS[2], "") or "")) or sid_orig
cf1_type = _get_cf_type_from_scene_file(csv_path, sid_cf1, "cf1")
cf2_type = _get_cf_type_from_scene_file(csv_path, sid_cf2, "cf2")
return (cf1_type, cf2_type)
def discover_csvs() -> list[Path]:
"""Find CSVs under output/, data/, and hf_dataset/ (recursive)."""
out: list[Path] = []
for base in (OUTPUT_DIR, DATA_DIR, HF_DATASET_DIR):
if base.exists():
out.extend(base.rglob("*.csv"))
return sorted(set(out), key=lambda p: str(p))
def _extract_upload_zip(zip_bytes: bytes) -> Path:
"""Extract zip to a temp directory; return path to that directory."""
root = Path(tempfile.mkdtemp(prefix="dataset_upload_"))
with zipfile.ZipFile(io.BytesIO(zip_bytes), "r") as z:
z.extractall(root)
return root
def _discover_csvs_in_dir(root: Path) -> list[Path]:
"""Find all CSVs under root (recursive)."""
return sorted(root.rglob("*.csv"), key=lambda p: str(p))
def _csv_from_upload(uploaded) -> tuple[Path, Path] | None:
"""
Process uploaded file (ZIP or CSV). Return (base_dir, csv_path) or None.
- ZIP: extract to temp, discover CSVs; base_dir = extract root, csv_path = chosen CSV.
- CSV: write to temp dir; base_dir = temp dir, csv_path = that CSV.
"""
fname = (uploaded.name or "").lower()
data = uploaded.read()
if fname.endswith(".zip"):
base = _extract_upload_zip(data)
csvs = _discover_csvs_in_dir(base)
if not csvs:
return None
# Prefer MMB-style CSV (name contains "question") if multiple
with_q = [p for p in csvs if "question" in p.name.lower()]
chosen = (with_q[0] if with_q else csvs[0])
return (base, chosen)
if fname.endswith(".csv"):
base = Path(tempfile.mkdtemp(prefix="dataset_upload_"))
path = base / (uploaded.name or "uploaded.csv")
path.write_bytes(data)
return (base, path)
return None
def detect_format(df: pd.DataFrame) -> str:
"""Return 'mmb_qa' | 'mmb_images' | 'mib' | 'unknown'."""
cols = set(df.columns)
has_mmb_images = all(c in cols for c in MMB_IMAGE_COLS)
if has_mmb_images:
if MMB_QA_COLS[0] in cols or "original_question" in cols:
return "mmb_qa"
return "mmb_images"
if "k" in cols and "f" in cols and "method" in cols:
return "mib"
if "dataset" in cols and "split" in cols and "count" in cols:
return "mib"
if "method" in cols and "task" in cols and "CPR" in cols:
return "mib"
return "unknown"
@st.cache_data
def load_csv(path: Path) -> pd.DataFrame:
return pd.read_csv(path)
QA_LABELS = {"original_question": "Original", "counterfactual1_question": "CF1", "counterfactual2_question": "CF2"}
# 3×3 grid: [image][question] -> CSV column name
ANSWER_GRID = [
["original_image_answer_to_original_question", "original_image_answer_to_cf1_question", "original_image_answer_to_cf2_question"],
["cf1_image_answer_to_original_question", "cf1_image_answer_to_cf1_question", "cf1_image_answer_to_cf2_question"],
["cf2_image_answer_to_original_question", "cf2_image_answer_to_cf1_question", "cf2_image_answer_to_cf2_question"],
]
def _render_image_cell(csv_path: Path | None, value, label: str):
"""Display image from file path (CSV/upload) or filename fallback."""
fname = str(value) if value is not None else ""
fp = resolve_image_path(csv_path, fname) if csv_path and fname else None
if fp:
st.image(str(fp), use_container_width=True, caption=label)
return
st.caption(label)
st.code(fname or "(no image)", language=None)
def _answers_as_grid(row, cols: list[str]) -> str | None:
"""Format 3×3 answer matrix as markdown table. Returns None if columns missing."""
def _cell(v):
return str(v).replace("|", "·").replace("\n", " ").strip()
grid = []
for r in range(3):
line = []
for c in range(3):
key = ANSWER_GRID[r][c]
if key not in cols or key not in row.index:
return None
line.append(_cell(row[key]))
grid.append(line)
header = "| | **Original Q** | **CF1 Q** | **CF2 Q** |"
sep = "| --- | --- | --- | --- |"
rows = [
"| **Original** | " + " | ".join(grid[0]) + " |",
"| **CF1** | " + " | ".join(grid[1]) + " |",
"| **CF2** | " + " | ".join(grid[2]) + " |",
]
return "\n".join([header, sep] + rows)
def render_mmb_qa(df: pd.DataFrame, csv_name: str, csv_path: Path | None = None):
st.subheader("MMB-style dataset: " + csv_name)
cols = df.columns.tolist()
tab_sets, tab_overview, tab_difficulty, tab_cf_types, tab_answers = st.tabs([
"Image sets", "Overview", "Difficulty & questions", "Counterfactual types", "Answer matrix",
])
with tab_sets:
st.markdown("**Scene sets**: original + 2 counterfactuals per row. Images and info below.")
include_answers = st.checkbox("Include answer matrix in each row", value=True, key="sets_answers")
if df.empty:
st.caption("No scene sets.")
else:
for i, (_, row) in enumerate(df.iterrows()):
sid = _row_scene_id(row, i)
cf1_t, cf2_t = get_cf_types_for_scene(csv_path, row) if csv_path else (None, None)
labels = [
"Original",
"Counterfactual 1" + (f" ({cf1_t})" if cf1_t else ""),
"Counterfactual 2" + (f" ({cf2_t})" if cf2_t else ""),
]
st.markdown(f"**{sid}** (row {i + 1})")
c1, c2, c3 = st.columns(3)
for j, (col_name, label) in enumerate([
(MMB_IMAGE_COLS[0], labels[0]),
(MMB_IMAGE_COLS[1], labels[1]),
(MMB_IMAGE_COLS[2], labels[2]),
]):
with (c1, c2, c3)[j]:
if col_name in row.index:
_render_image_cell(csv_path, row[col_name], label)
qa_lines = []
for qc, dc in zip(MMB_QA_COLS, MMB_DIFF_COLS):
if qc in row.index and dc in row.index:
lab = QA_LABELS.get(qc, qc)
qa_lines.append(f"**{lab}** ({row[dc]}): {row[qc]}")
for line in qa_lines:
st.markdown("- " + line)
if include_answers:
tbl = _answers_as_grid(row, cols)
if tbl:
st.markdown("**Answers** (image → question)")
st.markdown(tbl)
else:
ans_cols = [c for c in cols if "answer" in c.lower()]
if ans_cols:
st.json({c: str(row[c]) for c in ans_cols})
st.divider()
with tab_overview:
st.markdown("**Scene sets** (original + 2 counterfactuals per row).")
n = len(df)
st.metric("Scene sets", n)
if MMB_IMAGE_COLS[0] in cols:
st.caption("Columns: " + ", ".join(cols[:6]) + (" …" if len(cols) > 6 else ""))
fig = px.bar(
x=["Scene sets"],
y=[n],
title="Number of scene sets",
labels={"y": "count", "x": ""},
)
fig.update_layout(template="plotly_white", showlegend=False)
st.plotly_chart(fig, use_container_width=True)
with tab_difficulty:
diff_cols = [c for c in MMB_DIFF_COLS if c in cols]
if not diff_cols:
st.info("No difficulty columns in this CSV.")
else:
st.markdown("Question difficulty counts (original, CF1, CF2).")
label_map = {
"original_question_difficulty": "Original",
"counterfactual1_question_difficulty": "CF1",
"counterfactual2_question_difficulty": "CF2",
}
rows = []
for c in diff_cols:
vc = df[c].value_counts()
lab = label_map.get(c, c)
for lev, cnt in vc.items():
rows.append({"question": lab, "difficulty": str(lev), "count": int(cnt)})
diff_df = pd.DataFrame(rows)
fig = px.bar(
diff_df,
x="question",
y="count",
color="difficulty",
barmode="group",
title="Difficulty by question type",
color_discrete_map={"easy": "#2ecc71", "medium": "#f39c12", "hard": "#e74c3c"},
)
fig.update_layout(template="plotly_white", xaxis_tickangle=-20)
st.plotly_chart(fig, use_container_width=True)
with tab_cf_types:
cf_rows: list[dict] = []
for i, (_, row) in enumerate(df.iterrows()):
sid = _row_scene_id(row, i)
cf1_t, cf2_t = get_cf_types_for_scene(csv_path, row) if csv_path else (None, None)
cf_rows.append({"scene": sid, "CF1 type": cf1_t or "—", "CF2 type": cf2_t or "—"})
cf_df = pd.DataFrame(cf_rows)
if cf_df.empty:
st.caption("No scene sets.")
else:
st.markdown("**Counterfactual types** per scene (from `scenes/*_cf1.json`, `*_cf2.json`).")
st.dataframe(cf_df, use_container_width=True, hide_index=True)
# Counts for bar chart
flat: list[dict] = []
for r in cf_rows:
if r["CF1 type"] != "—":
flat.append({"slot": "CF1", "cf_type": r["CF1 type"]})
if r["CF2 type"] != "—":
flat.append({"slot": "CF2", "cf_type": r["CF2 type"]})
if flat:
flat_df = pd.DataFrame(flat)
agg = flat_df.groupby(["cf_type", "slot"]).size().reset_index(name="count")
fig = px.bar(
agg,
x="cf_type",
y="count",
color="slot",
barmode="group",
title="Counterfactual types by slot",
labels={"cf_type": "type"},
)
fig.update_layout(template="plotly_white", xaxis_tickangle=-45)
st.plotly_chart(fig, use_container_width=True)
else:
st.info("No `scenes/` folder or `cf_type` in scene JSONs. Add scenes next to the CSV.")
with tab_answers:
answer_cols = [c for c in cols if "answer" in c.lower()]
if not answer_cols:
st.info("No answer-matrix columns in this CSV.")
else:
st.markdown("Answer matrix: each image’s answer to each question (3×3 grid per set).")
for i, (_, row) in enumerate(df.iterrows()):
sid = _row_scene_id(row, i)
st.markdown(f"**{sid}** (row {i + 1})")
tbl = _answers_as_grid(row, cols)
if tbl:
st.markdown(tbl)
else:
st.json({c: str(row[c]) for c in answer_cols})
st.divider()
def render_mmb_images(df: pd.DataFrame, csv_name: str, csv_path: Path | None = None):
st.subheader("MMB-style (images only): " + csv_name)
n = len(df)
st.metric("Scene sets", n)
st.markdown("**Image sets** (images in each row below).")
if df.empty:
st.caption("No scene sets.")
else:
for i, (_, row) in enumerate(df.iterrows()):
sid = _row_scene_id(row, i)
cf1_t, cf2_t = get_cf_types_for_scene(csv_path, row) if csv_path else (None, None)
labels = [
"Original",
"CF1" + (f" ({cf1_t})" if cf1_t else ""),
"CF2" + (f" ({cf2_t})" if cf2_t else ""),
]
st.markdown(f"**{sid}** (row {i + 1})")
c1, c2, c3 = st.columns(3)
for j, (col_name, label) in enumerate([
(MMB_IMAGE_COLS[0], labels[0]),
(MMB_IMAGE_COLS[1], labels[1]),
(MMB_IMAGE_COLS[2], labels[2]),
]):
with (c1, c2, c3)[j]:
if col_name in row.index:
_render_image_cell(csv_path, row[col_name], label)
st.divider()
def main():
st.title("Dataset Visualizer")
st.caption("**Upload** your dataset (ZIP or CSV) or use **CSV files** from output/data.")
with st.sidebar:
source = st.radio("Data source", ["Upload dataset", "CSV files"], horizontal=True)
if source == "Upload dataset":
with st.sidebar:
st.header("Upload dataset")
uploaded = st.file_uploader(
"ZIP (CSV + images/ + optional scenes/) or CSV",
type=["zip", "csv"],
key="upload_dataset",
)
use_btn = st.button("Use this file", key="upload_use")
if st.session_state.get("upload_csv_path") is not None:
st.caption("Uploaded: " + (st.session_state.get("upload_name") or "dataset"))
clear_btn = st.button("Clear uploaded dataset", key="upload_clear")
else:
clear_btn = False
if clear_btn:
base = st.session_state.get("upload_base")
if base is not None:
try:
shutil.rmtree(base, ignore_errors=True)
except Exception:
pass
for k in ("upload_base", "upload_csv_path", "upload_name"):
st.session_state.pop(k, None)
st.rerun()
if use_btn and uploaded:
prev_base = st.session_state.get("upload_base")
if prev_base is not None:
try:
shutil.rmtree(prev_base, ignore_errors=True)
except Exception:
pass
with st.spinner("Processing upload…"):
out = _csv_from_upload(uploaded)
if out is None:
st.error("No CSV found in ZIP, or invalid upload.")
else:
base, csv_path = out
st.session_state["upload_base"] = base
st.session_state["upload_csv_path"] = csv_path
st.session_state["upload_name"] = uploaded.name
st.rerun()
csv_path = st.session_state.get("upload_csv_path")
if csv_path is None:
if uploaded and not use_btn:
st.info("Click **Use this file** to visualize the uploaded dataset.")
else:
st.info("Upload a **ZIP** (CSV + **images/** folder, optional **scenes/**) or **CSV** to visualize.")
with st.expander("Expected ZIP structure"):
st.code("""mydata.zip
image_mapping_with_questions.csv (or your CSV)
images/
scene_0001_original.png
scene_0001_cf1.png
...
scenes/ (optional, for counterfactual types)
scene_0001_cf1.json
...""", language="text")
return
path = Path(csv_path) if not isinstance(csv_path, Path) else csv_path
df = load_csv(path)
fmt = detect_format(df)
name = st.session_state.get("upload_name") or path.name
if fmt == "mmb_qa":
render_mmb_qa(df, name, csv_path=path)
elif fmt == "mmb_images":
render_mmb_images(df, name, csv_path=path)
elif fmt == "mib":
st.info("MIB-style CSV detected. Not supported for visualization.")
else:
st.warning("Unknown CSV format. Not supported for visualization.")
return
# CSV source
csv_options = discover_csvs()
if not csv_options:
st.info(
"No CSVs found in **output/** or **data/** or **hf_dataset/**. "
"Upload a dataset above (ZIP or CSV) or add CSV files to this repo and redeploy."
)
return
with st.sidebar:
st.header("CSV file")
rel_paths = []
for p in csv_options:
try:
rel_paths.append(str(p.relative_to(WORKSPACE_ROOT)))
except ValueError:
rel_paths.append(str(p))
chosen = st.selectbox("Choose CSV", options=rel_paths, format_func=lambda x: x)
path = next(p for p in csv_options if (str(p.relative_to(WORKSPACE_ROOT)) == chosen or str(p) == chosen))
df = load_csv(path)
fmt = detect_format(df)
csv_name = path.name
if fmt == "mmb_qa":
render_mmb_qa(df, csv_name, csv_path=path)
elif fmt == "mmb_images":
render_mmb_images(df, csv_name, csv_path=path)
elif fmt == "mib":
st.info("MIB-style CSV detected. Not supported for visualization.")
else:
st.warning("Unknown CSV format. Not supported for visualization.")
if __name__ == "__main__":
main()
|