File size: 12,884 Bytes
ce847d4 |
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 |
"""OCR engine — ctypes wrapper for Windows 11 SnippingTool's oneocr.dll.
Provides offline OCR capability using Microsoft's AI model from Snipping Tool.
Requires oneocr.dll, oneocr.onemodel, and onnxruntime.dll in ocr_data/ directory.
Usage:
from src.services.ocr.engine import OcrEngine
engine = OcrEngine()
result = engine.recognize_pil(pil_image)
print(result.text)
"""
from __future__ import annotations
import ctypes
import os
from contextlib import contextmanager
from ctypes import (
POINTER,
Structure,
byref,
c_char_p,
c_float,
c_int32,
c_int64,
c_ubyte,
)
from pathlib import Path
from typing import TYPE_CHECKING
from ocr.models import BoundingRect, OcrLine, OcrResult, OcrWord
# Constants (previously imported from src.config.constants)
OCR_DLL_NAME = "oneocr.dll"
OCR_MODEL_NAME = "oneocr.onemodel"
OCR_MODEL_KEY = b'kj)TGtrK>f]b[Piow.gU+nC@s""""""4'
OCR_MAX_LINES = 200
if TYPE_CHECKING:
from PIL import Image
c_int64_p = POINTER(c_int64)
c_float_p = POINTER(c_float)
c_ubyte_p = POINTER(c_ubyte)
class _ImageStructure(Structure):
"""Image data structure for oneocr.dll (CV_8UC4 format)."""
_fields_ = [
("type", c_int32),
("width", c_int32),
("height", c_int32),
("_reserved", c_int32),
("step_size", c_int64),
("data_ptr", c_ubyte_p),
]
class _BoundingBox(Structure):
"""Bounding box coordinates from DLL."""
_fields_ = [
("x1", c_float), ("y1", c_float),
("x2", c_float), ("y2", c_float),
("x3", c_float), ("y3", c_float),
("x4", c_float), ("y4", c_float),
]
_BoundingBox_p = POINTER(_BoundingBox)
# DLL function signatures: (name, argtypes, restype)
_DLL_FUNCTIONS: list[tuple[str, list[type], type | None]] = [
("CreateOcrInitOptions", [c_int64_p], c_int64),
("OcrInitOptionsSetUseModelDelayLoad", [c_int64, ctypes.c_char], c_int64),
("CreateOcrPipeline", [c_char_p, c_char_p, c_int64, c_int64_p], c_int64),
("CreateOcrProcessOptions", [c_int64_p], c_int64),
("OcrProcessOptionsSetMaxRecognitionLineCount", [c_int64, c_int64], c_int64),
("RunOcrPipeline", [c_int64, POINTER(_ImageStructure), c_int64, c_int64_p], c_int64),
("GetImageAngle", [c_int64, c_float_p], c_int64),
("GetOcrLineCount", [c_int64, c_int64_p], c_int64),
("GetOcrLine", [c_int64, c_int64, c_int64_p], c_int64),
("GetOcrLineContent", [c_int64, POINTER(c_char_p)], c_int64),
("GetOcrLineBoundingBox", [c_int64, POINTER(_BoundingBox_p)], c_int64),
("GetOcrLineWordCount", [c_int64, c_int64_p], c_int64),
("GetOcrWord", [c_int64, c_int64, c_int64_p], c_int64),
("GetOcrWordContent", [c_int64, POINTER(c_char_p)], c_int64),
("GetOcrWordBoundingBox", [c_int64, POINTER(_BoundingBox_p)], c_int64),
("GetOcrWordConfidence", [c_int64, c_float_p], c_int64),
("ReleaseOcrResult", [c_int64], None),
("ReleaseOcrInitOptions", [c_int64], None),
("ReleaseOcrPipeline", [c_int64], None),
("ReleaseOcrProcessOptions", [c_int64], None),
]
@contextmanager
def _suppress_output():
"""Suppress stdout/stderr during DLL initialization (it prints to console)."""
devnull = os.open(os.devnull, os.O_WRONLY)
original_stdout = os.dup(1)
original_stderr = os.dup(2)
os.dup2(devnull, 1)
os.dup2(devnull, 2)
try:
yield
finally:
os.dup2(original_stdout, 1)
os.dup2(original_stderr, 2)
os.close(original_stdout)
os.close(original_stderr)
os.close(devnull)
class OcrEngine:
"""Offline OCR engine using Windows 11 SnippingTool's oneocr.dll.
Args:
ocr_data_dir: Path to directory containing oneocr.dll, oneocr.onemodel, onnxruntime.dll.
Defaults to PROJECT_ROOT/ocr_data/.
"""
def __init__(self, ocr_data_dir: str | Path | None = None) -> None:
if ocr_data_dir is None:
ocr_data_dir = Path(__file__).resolve().parent.parent / "ocr_data"
self._data_dir = str(Path(ocr_data_dir).resolve())
self._dll: ctypes.WinDLL | None = None # type: ignore[name-defined]
self._init_options = c_int64()
self._pipeline = c_int64()
self._process_options = c_int64()
self._load_dll()
self._initialize_pipeline()
def __del__(self) -> None:
if self._dll:
try:
self._dll.ReleaseOcrProcessOptions(self._process_options)
self._dll.ReleaseOcrPipeline(self._pipeline)
self._dll.ReleaseOcrInitOptions(self._init_options)
except Exception:
pass
# --- Public API ---
def recognize_pil(self, image: Image.Image) -> OcrResult:
"""Run OCR on a PIL Image.
Args:
image: PIL Image object (any mode — will be converted to RGBA/BGRA).
Returns:
OcrResult with recognized text, lines, words, and confidence values.
"""
if any(x < 50 or x > 10000 for x in image.size):
return OcrResult(error="Unsupported image size (must be 50-10000px)")
if image.mode != "RGBA":
image = image.convert("RGBA")
# Convert RGB(A) → BGRA (DLL expects BGRA channel order)
r, g, b, a = image.split()
from PIL import Image as PILImage
bgra_image = PILImage.merge("RGBA", (b, g, r, a))
return self._process_image(
width=bgra_image.width,
height=bgra_image.height,
step=bgra_image.width * 4,
data=bgra_image.tobytes(),
)
def recognize_bytes(self, image_bytes: bytes) -> OcrResult:
"""Run OCR on raw image bytes (PNG/JPEG/etc).
Args:
image_bytes: Raw image file bytes.
Returns:
OcrResult.
"""
from io import BytesIO
from PIL import Image
img = Image.open(BytesIO(image_bytes))
return self.recognize_pil(img)
# --- Internal ---
def _load_dll(self) -> None:
"""Load oneocr.dll and bind function signatures."""
try:
kernel32 = ctypes.WinDLL("kernel32", use_last_error=True) # type: ignore[attr-defined]
if hasattr(kernel32, "SetDllDirectoryW"):
kernel32.SetDllDirectoryW(self._data_dir)
dll_path = os.path.join(self._data_dir, OCR_DLL_NAME)
if not os.path.exists(dll_path):
raise FileNotFoundError(f"OCR DLL not found: {dll_path}")
self._dll = ctypes.WinDLL(dll_path) # type: ignore[attr-defined]
for name, argtypes, restype in _DLL_FUNCTIONS:
func = getattr(self._dll, name)
func.argtypes = argtypes
func.restype = restype
except (OSError, RuntimeError) as e:
raise RuntimeError(f"Failed to load OCR DLL from {self._data_dir}: {e}") from e
def _initialize_pipeline(self) -> None:
"""Create OCR init options, pipeline, and process options."""
assert self._dll is not None
# Init options
self._check(
self._dll.CreateOcrInitOptions(byref(self._init_options)),
"CreateOcrInitOptions failed",
)
self._check(
self._dll.OcrInitOptionsSetUseModelDelayLoad(self._init_options, 0),
"OcrInitOptionsSetUseModelDelayLoad failed",
)
# Pipeline (loads AI model — suppress DLL stdout noise)
model_path = os.path.join(self._data_dir, OCR_MODEL_NAME)
if not os.path.exists(model_path):
raise FileNotFoundError(f"OCR model not found: {model_path}")
model_buf = ctypes.create_string_buffer(model_path.encode())
key_buf = ctypes.create_string_buffer(OCR_MODEL_KEY)
with _suppress_output():
result = self._dll.CreateOcrPipeline(
model_buf, key_buf, self._init_options, byref(self._pipeline)
)
self._check(result, "CreateOcrPipeline failed (wrong key or corrupted model?)")
# Process options
self._check(
self._dll.CreateOcrProcessOptions(byref(self._process_options)),
"CreateOcrProcessOptions failed",
)
self._check(
self._dll.OcrProcessOptionsSetMaxRecognitionLineCount(
self._process_options, OCR_MAX_LINES
),
"OcrProcessOptionsSetMaxRecognitionLineCount failed",
)
def _process_image(self, width: int, height: int, step: int, data: bytes) -> OcrResult:
"""Create image structure and run OCR pipeline."""
assert self._dll is not None
data_ptr = (c_ubyte * len(data)).from_buffer_copy(data)
img_struct = _ImageStructure(
type=3,
width=width,
height=height,
_reserved=0,
step_size=step,
data_ptr=data_ptr,
)
ocr_result = c_int64()
if self._dll.RunOcrPipeline(
self._pipeline, byref(img_struct), self._process_options, byref(ocr_result)
) != 0:
return OcrResult(error="RunOcrPipeline returned non-zero")
parsed = self._parse_results(ocr_result)
self._dll.ReleaseOcrResult(ocr_result)
return parsed
def _parse_results(self, ocr_result: c_int64) -> OcrResult:
"""Extract text, lines, words from DLL result handle."""
assert self._dll is not None
line_count = c_int64()
if self._dll.GetOcrLineCount(ocr_result, byref(line_count)) != 0:
return OcrResult(error="GetOcrLineCount failed")
lines: list[OcrLine] = []
for idx in range(line_count.value):
line = self._parse_line(ocr_result, idx)
if line:
lines.append(line)
# Text angle
text_angle_val = c_float()
text_angle: float | None = None
if self._dll.GetImageAngle(ocr_result, byref(text_angle_val)) == 0:
text_angle = text_angle_val.value
full_text = "\n".join(line.text for line in lines if line.text)
return OcrResult(text=full_text, text_angle=text_angle, lines=lines)
def _parse_line(self, ocr_result: c_int64, line_index: int) -> OcrLine | None:
"""Parse a single line from OCR result."""
assert self._dll is not None
line_handle = c_int64()
if self._dll.GetOcrLine(ocr_result, line_index, byref(line_handle)) != 0:
return None
if not line_handle.value:
return None
# Line text
content = c_char_p()
line_text = ""
if self._dll.GetOcrLineContent(line_handle, byref(content)) == 0 and content.value:
line_text = content.value.decode("utf-8", errors="ignore")
# Line bounding box
line_bbox = self._get_bbox(line_handle, self._dll.GetOcrLineBoundingBox)
# Words
word_count = c_int64()
words: list[OcrWord] = []
if self._dll.GetOcrLineWordCount(line_handle, byref(word_count)) == 0:
for wi in range(word_count.value):
word = self._parse_word(line_handle, wi)
if word:
words.append(word)
return OcrLine(text=line_text, bounding_rect=line_bbox, words=words)
def _parse_word(self, line_handle: c_int64, word_index: int) -> OcrWord | None:
"""Parse a single word."""
assert self._dll is not None
word_handle = c_int64()
if self._dll.GetOcrWord(line_handle, word_index, byref(word_handle)) != 0:
return None
# Word text
content = c_char_p()
word_text = ""
if self._dll.GetOcrWordContent(word_handle, byref(content)) == 0 and content.value:
word_text = content.value.decode("utf-8", errors="ignore")
# Word bounding box
word_bbox = self._get_bbox(word_handle, self._dll.GetOcrWordBoundingBox)
# Word confidence
confidence_val = c_float()
confidence = 0.0
if self._dll.GetOcrWordConfidence(word_handle, byref(confidence_val)) == 0:
confidence = confidence_val.value
return OcrWord(text=word_text, bounding_rect=word_bbox, confidence=confidence)
@staticmethod
def _get_bbox(handle: c_int64, bbox_fn: object) -> BoundingRect | None:
"""Extract bounding box from a handle."""
bbox_ptr = _BoundingBox_p()
if bbox_fn(handle, byref(bbox_ptr)) == 0 and bbox_ptr: # type: ignore[operator]
bb = bbox_ptr.contents
return BoundingRect(
x1=bb.x1, y1=bb.y1, x2=bb.x2, y2=bb.y2,
x3=bb.x3, y3=bb.y3, x4=bb.x4, y4=bb.y4,
)
return None
@staticmethod
def _check(result_code: int, msg: str) -> None:
if result_code != 0:
raise RuntimeError(f"{msg} (code: {result_code})")
|