File size: 9,374 Bytes
44a7ed3 | 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 | #!/usr/bin/env python3
"""Build MVEB Product1m test subset from downloaded images and test_split.json.
Input:
scripts/Product1m/test_split.json
{
"test": [{"id": "...", "line_number": N}, ...],
"gallery": [{"id": "...", "line_number": N}, ...]
}
source/product1m/{test,gallery}/*.jpg
downloads/product1m/Product1M/Poudct1M/product1m_{test,gallery}_ossurl_v2.txt
Output (when executed from MVEB root):
./MVEB-test/Product1m/{query.parquet,candidate.parquet,media-*.parquet,README.md}
"""
from __future__ import annotations
import argparse
import json
import shutil
import sys
from collections import defaultdict
from pathlib import Path
from typing import Dict, List, Sequence, Tuple
SCRIPT_DIR = Path(__file__).resolve().parent
SCRIPTS_ROOT = SCRIPT_DIR.parent
ROOT_DIR = SCRIPT_DIR.parent.parent.parent
if str(SCRIPTS_ROOT) not in sys.path:
sys.path.insert(0, str(SCRIPTS_ROOT))
from pack_media_parquet import pack_dataset_with_media, resolve_split_output_dir
QUERY_INSTRUCTION = "You are a helpful assistant."
QUERY_TEXT = "Represent the product in the given image."
CANDIDATE_INSTRUCTION = "You are a helpful assistant."
CANDIDATE_TEXT = "Represent the product in the given image."
TXT_FILES = {
"test": "product1m_test_ossurl_v2.txt",
"gallery": "product1m_gallery_ossurl_v2.txt",
}
def _load_split(path: Path) -> dict:
with path.open("r", encoding="utf-8") as f:
data = json.load(f)
for key in ("test", "gallery"):
if key not in data or not isinstance(data[key], list):
raise ValueError(f"split json must contain key {key!r} with a list value")
return data
def _read_lines_at(txt_path: Path, line_numbers: Sequence[int]) -> Dict[int, str]:
wanted = set(line_numbers)
max_line = max(line_numbers)
found: Dict[int, str] = {}
with txt_path.open("r", encoding="utf-8") as f:
for lineno, line in enumerate(f, 1):
if lineno in wanted:
found[lineno] = line.strip()
if lineno >= max_line and len(found) == len(wanted):
break
missing = sorted(wanted - found.keys())
if missing:
raise ValueError(f"Missing lines in {txt_path}: {missing[:10]}")
return found
def _parse_record(line: str) -> Tuple[str, List[str]]:
"""Parse Product1M txt line -> (image_id, product_ids)."""
parts = line.split("#####")
if len(parts) < 4:
raise ValueError(f"Invalid Product1M line (<4 fields): {line[:120]}")
image_id = parts[0]
# Official converter uses split("####") then strip('#') on product tokens.
infos = line.split("####")
product_ids = [item.strip("#") for item in infos[-1].split(";") if item.strip("#")]
return image_id, product_ids
def _make_annotations(
split_data: dict,
anno_dir: Path,
image_root: Path,
) -> Tuple[List[dict], List[dict]]:
"""Build query/candidate rows aligned with converter/product1m.py test logic.
- candidates: all gallery images in test_split.json
- queries: all test images that have >=1 matching gallery product_id
- pos_ids: gallery candidate ids sharing any product_id with the query
"""
gallery_entries = split_data["gallery"]
test_entries = split_data["test"]
gallery_lines = _read_lines_at(
anno_dir / TXT_FILES["gallery"],
[item["line_number"] for item in gallery_entries],
)
test_lines = _read_lines_at(
anno_dir / TXT_FILES["test"],
[item["line_number"] for item in test_entries],
)
productid2candidate_ids: Dict[str, List[str]] = defaultdict(list)
candidate_rows: List[dict] = []
for idx, item in enumerate(gallery_entries):
line = gallery_lines[item["line_number"]]
image_id, product_ids = _parse_record(line)
if image_id != item["id"]:
raise ValueError(
f"Gallery ID mismatch at line {item['line_number']}: "
f"split={item['id']}, txt={image_id}"
)
rel_path = f"gallery/{image_id}.jpg"
abs_path = image_root / rel_path
if not abs_path.is_file():
raise FileNotFoundError(f"Missing gallery image: {abs_path}")
candidate_rows.append(
{
"id": rel_path,
"image_path": rel_path,
"instruction": CANDIDATE_INSTRUCTION,
"text": CANDIDATE_TEXT,
}
)
for product_id in product_ids:
productid2candidate_ids[product_id].append(rel_path)
query_rows: List[dict] = []
skipped = 0
for item in test_entries:
line = test_lines[item["line_number"]]
image_id, product_ids = _parse_record(line)
if image_id != item["id"]:
raise ValueError(
f"Test ID mismatch at line {item['line_number']}: "
f"split={item['id']}, txt={image_id}"
)
rel_path = f"test/{image_id}.jpg"
abs_path = image_root / rel_path
if not abs_path.is_file():
raise FileNotFoundError(f"Missing test image: {abs_path}")
pos_ids: List[str] = []
seen = set()
for product_id in product_ids:
for cid in productid2candidate_ids.get(product_id, []):
if cid not in seen:
seen.add(cid)
pos_ids.append(cid)
if not pos_ids:
skipped += 1
continue
query_rows.append(
{
"id": str(len(query_rows)),
"image_path": rel_path,
"instruction": QUERY_INSTRUCTION,
"text": QUERY_TEXT,
"pos_ids": pos_ids,
}
)
if skipped:
print(f"[warn] skipped {skipped} test queries with empty pos_ids")
if not query_rows:
raise ValueError("No valid query annotations were built")
if not candidate_rows:
raise ValueError("No candidate annotations were built")
return query_rows, candidate_rows
def _process_test(
split_data: dict,
anno_dir: Path,
image_root: Path,
output_root: Path,
overwrite: bool,
media_rows_per_shard: int,
row_group_size: int,
num_workers: int,
) -> None:
out_dir = resolve_split_output_dir(output_root, "test", "Product1m")
if overwrite and out_dir.exists():
shutil.rmtree(out_dir)
out_dir.mkdir(parents=True, exist_ok=True)
query_rows, candidate_rows = _make_annotations(split_data, anno_dir, image_root)
stats = pack_dataset_with_media(
query_annotations=query_rows,
candidate_annotations=candidate_rows,
image_dir=str(image_root),
output_dir=str(out_dir),
media_rows_per_shard=media_rows_per_shard,
row_group_size=row_group_size,
num_workers=num_workers,
dataset_name="Product1m",
data_split="test",
write_subset_readme=True,
show_progress=True,
)
print(
f"[test] done: media={stats['num_media']}, "
f"query={stats['num_query']}, candidate={stats['num_candidate']}, "
f"shards={stats['num_shards']} -> {out_dir}"
)
def main() -> None:
parser = argparse.ArgumentParser(
description="Process Product1m test_split.json to MVEB parquet format."
)
parser.add_argument(
"--split-json",
type=Path,
default=SCRIPT_DIR / "test_split.json",
help="JSON containing test/gallery id + line_number entries.",
)
parser.add_argument(
"--image-root",
type=Path,
default=ROOT_DIR / "source" / "product1m",
help="Root directory containing test/ and gallery/ images.",
)
parser.add_argument(
"--repo-dir",
type=Path,
default=ROOT_DIR / "downloads" / "product1m" / "Product1M",
help="Cloned Product1M github repo.",
)
parser.add_argument(
"--output-root",
type=Path,
default=ROOT_DIR,
help="Output MVEB root directory (contains test/).",
)
parser.add_argument(
"--overwrite",
action="store_true",
help="Delete existing output split dir before writing.",
)
parser.add_argument("--media-rows-per-shard", type=int, default=5000)
parser.add_argument("--row-group-size", type=int, default=100)
parser.add_argument("--num-workers", type=int, default=1)
args = parser.parse_args()
if not args.split_json.exists():
raise FileNotFoundError(f"split json not found: {args.split_json}")
if not args.image_root.exists():
raise FileNotFoundError(f"image root not found: {args.image_root}")
anno_dir = args.repo_dir / "Poudct1M"
if not anno_dir.exists():
raise FileNotFoundError(f"anno dir not found: {anno_dir}")
for name in TXT_FILES.values():
txt_path = anno_dir / name
if not txt_path.exists():
raise FileNotFoundError(f"annotation txt not found: {txt_path}")
split_data = _load_split(args.split_json)
_process_test(
split_data=split_data,
anno_dir=anno_dir,
image_root=args.image_root,
output_root=args.output_root,
overwrite=args.overwrite,
media_rows_per_shard=args.media_rows_per_shard,
row_group_size=args.row_group_size,
num_workers=args.num_workers,
)
if __name__ == "__main__":
main()
|