File size: 5,356 Bytes
dadf189 | 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 | from __future__ import annotations
from dataclasses import dataclass, field
from pathlib import Path
from typing import Iterable
import torch
from PIL import Image
from torchvision import transforms
from .base import FingerprintSample
IMAGE_EXTS = {".png", ".jpg", ".jpeg", ".bmp", ".tif", ".tiff"}
@dataclass
class FVCPaths:
"""Root directories for FVC years to include.
Each entry is the 'Dbs' folder (or parent of Dbs).
Example:
root_fvc2002 = "/path/to/FVC_Dataset/FVC2002"
root_fvc2004 = "/path/to/FVC_Dataset/FVC2004"
Leave empty string to skip a year.
"""
root_fvc2002: str = ""
root_fvc2004: str = ""
root_fvc2000: str = ""
extra_roots: list[str] = field(default_factory=list)
class FVCLoader:
"""Loader for FVC 2000/2002/2004 datasets.
Directory layout (standard FVC):
<year_root>/Dbs/<DbN_a|DbN_b>/<subject>_<impression>.tif
Metadata:
identity_id — "<year>_<db>_<subject_id>" e.g. "2002_db1_042"
8 impressions per subject → real quality variation for L_mat
finger_id — "f01" for all FVC (one finger per subject in competition protocol)
sensor_id — "fvc<year>_db<N>" e.g. "fvc2002_db1"
dataset — "fvc<year>"
Subsets _a and _b belong to the same sensor_id — they are train/test splits
of the same capture device, not different devices.
"""
def __init__(self, image_size: int = 224):
self.transform = transforms.Compose(
[
transforms.Resize((image_size, image_size)),
transforms.ToTensor(),
]
)
def discover(self, paths: FVCPaths) -> list[dict[str, str]]:
records: list[dict[str, str]] = []
year_roots = [
(paths.root_fvc2000, "2000"),
(paths.root_fvc2002, "2002"),
(paths.root_fvc2004, "2004"),
]
for extra in paths.extra_roots:
year_roots.append((extra, "extra"))
for root_str, year in year_roots:
if not root_str:
continue
root = Path(root_str)
dbs_dir = root / "Dbs" if (root / "Dbs").exists() else root
records.extend(self._scan_dbs(dbs_dir, year))
return records
def iter_samples(
self, records: Iterable[dict[str, str]]
) -> Iterable[FingerprintSample]:
for rec in records:
image = Image.open(rec["image_path"]).convert("L")
tensor = self.transform(image)
yield {
"image": tensor,
"identity_id": rec["identity_id"],
"finger_id": rec["finger_id"],
"sensor_id": rec["sensor_id"],
"dataset": rec["dataset"],
"image_path": rec["image_path"],
}
def _scan_dbs(self, dbs_root: Path, year: str) -> list[dict[str, str]]:
if not dbs_root.exists():
return []
records: list[dict[str, str]] = []
for db_dir in sorted(dbs_root.iterdir()):
if not db_dir.is_dir():
continue
db_name = db_dir.name.lower() # e.g. "db1_a", "DB1_A"
# Normalise: "db1_a" or "DB1_A" → db_id = "db1"
db_id = self._parse_db_id(db_name)
if db_id is None:
continue
sensor_id = f"fvc{year}_{db_id}"
dataset = f"fvc{year}"
for img_path in sorted(db_dir.iterdir()):
if not img_path.is_file() or img_path.suffix.lower() not in IMAGE_EXTS:
continue
meta = self._parse_filename(img_path, sensor_id, dataset, year, db_id)
if meta is not None:
records.append(meta)
return records
@staticmethod
def _parse_db_id(db_dir_name: str) -> str | None:
"""'Db1_a', 'DB2_B', 'db3_a' → 'db1', 'db2', 'db3'."""
name = db_dir_name.lower().strip()
# Strip trailing subset suffix (_a, _b, _A, _B)
for suffix in ("_a", "_b"):
if name.endswith(suffix):
name = name[: -len(suffix)]
break
# Accept names like 'db1', 'db2', 'db3', 'db4'
if name.startswith("db") and name[2:].isdigit():
return name
return None
@staticmethod
def _parse_filename(
img_path: Path,
sensor_id: str,
dataset: str,
year: str,
db_id: str,
) -> dict[str, str] | None:
"""Parse '<subject_id>_<impression_id>.tif' → metadata record."""
stem = img_path.stem # e.g. "042_3" or "100_8"
parts = stem.split("_")
if len(parts) != 2:
return None
subject_str, impression_str = parts
if not subject_str.isdigit() or not impression_str.isdigit():
return None
subject_id = int(subject_str)
impression_id = int(impression_str)
# Unique identity key: year + db + subject_id
identity_id = f"{year}_{db_id}_{subject_id:04d}"
return {
"identity_id": identity_id,
"finger_id": "f01", # FVC: one finger per subject
"impression_id": str(impression_id),
"sensor_id": sensor_id,
"dataset": dataset,
"image_path": str(img_path),
}
|