File size: 18,905 Bytes
f6f8d06 |
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 |
import re
import numpy as np
import time
import cv2
import os
from typing import List
import ctypes
from ctypes import (
Structure, byref, POINTER, c_int64, c_int32, c_float, c_ubyte, c_char, c_char_p
)
from PIL import Image as PilImage
from contextlib import contextmanager
import logging
from .base import register_OCR, OCRBase, TextBlock
ONE_OCR_PATH = os.path.join("data", "models", "one-ocr")
MODEL_NAME = "oneocr.onemodel"
DLL_NAME = "oneocr.dll"
MODEL_KEY = b'kj)TGtrK>f]b[Piow.gU+nC@s""""""4'
MIN_DIM_SIZE = 56 # Minimum height *and* width for padding
c_int64_p = POINTER(c_int64)
c_float_p = POINTER(c_float)
c_ubyte_p = POINTER(c_ubyte)
class ImageStructure(Structure):
_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):
_fields_ = [("x1", c_float), ("y1", c_float),
("x2", c_float), ("y2", c_float)]
BoundingBox_p = POINTER(BoundingBox)
DLL_FUNCTIONS = [ # Function definitions for the DLL
("CreateOcrInitOptions", [c_int64_p], c_int64), ("OcrInitOptionsSetUseModelDelayLoad", [
c_int64, 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(): # Context manager to suppress stdout/stderr, useful for noisy DLL init
devnull = os.open(os.devnull, os.O_WRONLY)
original_stdout, original_stderr = os.dup(1), 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:
def __init__(self, config_dir, logger=None):
self.ocr_dll = self.init_options = self.pipeline = self.process_options = None
self.config_dir = config_dir
self.model_path = os.path.join(self.config_dir, MODEL_NAME)
self.dll_path = os.path.join(self.config_dir, DLL_NAME)
self.logger = logger or logging.getLogger(self.__class__.__name__)
self.empty_result = {"text": "", "text_angle": None, "lines": []}
try:
self._load_and_bind_dll()
self.init_options = self._create_init_options()
self.pipeline = self._create_pipeline()
self.process_options = self._create_process_options()
if self.logger:
self.logger.debug("OcrEngine initialized")
except Exception as e:
raise e # Error logged internally
def _load_and_bind_dll(self):
try:
if os.name == 'nt':
k32 = ctypes.WinDLL("kernel32", use_last_error=True)
if hasattr(k32, "SetDllDirectoryW"):
k32.SetDllDirectoryW(str(self.config_dir))
self.ocr_dll = ctypes.WinDLL(
str(self.dll_path), use_last_error=True)
for name, argtypes, restype in DLL_FUNCTIONS:
try:
func = getattr(self.ocr_dll, name)
func.argtypes = argtypes
func.restype = restype
except AttributeError as e:
raise RuntimeError(f"Missing DLL func: {name}") from e
except (OSError, RuntimeError, AttributeError) as e:
code = ctypes.get_last_error() if os.name == "nt" else 0
msg = f"Failed load/bind DLL ({self.dll_path}) from {self.config_dir}. Code: {code}. Error: {e}"
if self.logger:
self.logger.error(msg)
raise RuntimeError(msg) from e
def _check_dll_result(self, result_code, error_message):
if result_code != 0:
raise RuntimeError(f"{error_message} (Native Code: {result_code})")
def _create_init_options(self):
h = c_int64()
self._check_dll_result(self.ocr_dll.CreateOcrInitOptions(
byref(h)), "Init options create failed")
self._check_dll_result(self.ocr_dll.OcrInitOptionsSetUseModelDelayLoad(
h, 0), "Model load config failed")
return h
def _create_pipeline(self):
mb = ctypes.create_string_buffer(self.model_path.encode("utf-8"))
kb = ctypes.create_string_buffer(MODEL_KEY)
h = c_int64()
with suppress_output():
self._check_dll_result(self.ocr_dll.CreateOcrPipeline(
mb, kb, self.init_options, byref(h)), f"Pipeline create failed ({self.model_path})")
return h
def _create_process_options(self):
h = c_int64()
self._check_dll_result(self.ocr_dll.CreateOcrProcessOptions(
byref(h)), "Process options create failed")
self._check_dll_result(self.ocr_dll.OcrProcessOptionsSetMaxRecognitionLineCount(
h, 1000), "Line count config failed")
return h
def recognize_pil(self, image: PilImage.Image):
if image.mode != 'RGBA':
image = image.convert('RGBA')
return self._process_image(cols=image.width, rows=image.height, step=image.width*4, data=image.tobytes())
def _process_image(self, cols, rows, step, data):
dp = ctypes.cast(data, c_ubyte_p) if not isinstance(
data, bytes) else (c_ubyte*len(data)).from_buffer_copy(data)
img_struct = ImageStructure(3, cols, rows, 0, step, dp)
return self._perform_ocr(img_struct)
def _perform_ocr(self, image_struct: ImageStructure):
res_h = c_int64()
code = self.ocr_dll.RunOcrPipeline(self.pipeline, byref(
image_struct), self.process_options, byref(res_h))
if code != 0:
# Let the caller (OCROneAPI) handle specific error code interpretation
# This will raise RuntimeError
self._check_dll_result(code, "RunOcrPipeline failed")
# Alternative: log here and return empty, but raising is cleaner for caller
# if self.logger: self.logger.warning(f"RunOcrPipeline failed: {code}")
# return self.empty_result
parsed = self._parse_ocr_results(res_h.value)
self.ocr_dll.ReleaseOcrResult(res_h)
return parsed
def _parse_ocr_results(self, res_h: int):
lc = c_int64()
if self.ocr_dll.GetOcrLineCount(res_h, byref(lc)) != 0:
if self.logger:
self.logger.warning("Failed get line count")
return self.empty_result
lines = self._get_lines(res_h, lc.value)
angle = self._get_text_angle(res_h)
return {"text": None, "text_angle": angle, "lines": lines}
def _get_text_angle(self, res_h: int): a = c_float(
); return a.value if self.ocr_dll.GetImageAngle(res_h, byref(a)) == 0 else None
def _get_lines(self, res_h: int, n: int): return [
self._process_line(res_h, i) for i in range(n)]
def _process_line(self, res_h: int, idx: int):
lh = c_int64()
if self.ocr_dll.GetOcrLine(res_h, idx, byref(lh)) != 0:
if self.logger:
self.logger.warning(f"Failed get line handle {idx}")
return {"text": None, "bounding_rect": None, "words": []}
lhv = lh.value
text = self._get_text(lhv, self.ocr_dll.GetOcrLineContent)
bbox = self._get_bounding_box(lhv, self.ocr_dll.GetOcrLineBoundingBox)
words = self._get_words(lhv)
return {"text": text, "bounding_rect": bbox, "words": words}
def _get_words(self, line_h: int):
wc = c_int64()
if self.ocr_dll.GetOcrLineWordCount(line_h, byref(wc)) != 0:
return []
return [self._process_word(line_h, i) for i in range(wc.value)]
def _process_word(self, line_h: int, idx: int):
wh = c_int64()
if self.ocr_dll.GetOcrWord(line_h, idx, byref(wh)) != 0:
if self.logger:
self.logger.warning(f"Failed get word handle {idx}")
return {"text": None, "bounding_rect": None, "confidence": None}
whv = wh.value
text = self._get_text(whv, self.ocr_dll.GetOcrWordContent)
bbox = self._get_bounding_box(whv, self.ocr_dll.GetOcrWordBoundingBox)
conf = self._get_word_confidence(whv)
return {"text": text, "bounding_rect": bbox, "confidence": conf}
def _get_text(self, handle: int, func):
content = c_char_p()
if func(handle, byref(content)) == 0 and content.value:
try:
return content.value.decode("utf-8", errors="ignore")
except Exception as e:
if self.logger:
self.logger.error(f"Error decoding text: {e}")
return None
def _get_word_confidence(self, word_h: int): c = c_float(
); return c.value if self.ocr_dll.GetOcrWordConfidence(word_h, byref(c)) == 0 else None
def _get_bounding_box(self, handle: int, func):
bp = BoundingBox_p()
if func(handle, byref(bp)) == 0 and bp:
b = bp.contents
return {"x1": b.x1, "y1": b.y1, "x2": b.x2, "y2": b.y2}
return None
@register_OCR("one_ocr")
class OCROneAPI(OCRBase):
params = {
"expand_small_blocks": {"type": "checkbox", "value": True, "description": f"Expand image width/height if < {MIN_DIM_SIZE}px before recognition by padding (helps with small images)"},
"newline_handling": {"type": "selector", "options": ["preserve", "remove"], "value": "preserve", "description": "Newline char handling (preserve/remove)"},
"reverse_line_order": {"type": "checkbox", "value": False, "description": "Reverse line order (for vertical CJK)"},
"no_uppercase": {"type": "checkbox", "value": False, "description": "Convert text to lowercase (except sentence start)"},
"description": "Local OCR using OneOCR library (Windows Only)",
}
@property
def expand_small_blocks(self): v = self.get_param_value(
"expand_small_blocks"); return bool(v) if v is not None else True
@property
def newline_handling(self): return self.get_param_value(
"newline_handling") or "preserve"
@property
def no_uppercase(self): return bool(self.get_param_value("no_uppercase"))
@property
def reverse_line_order(self): return bool(
self.get_param_value("reverse_line_order"))
def __init__(self, **params) -> None:
super().__init__(**params)
self.engine = None
self.available = False
self.config_dir = ONE_OCR_PATH
if os.name != "nt":
if self.logger:
self.logger.warning("OneOCR is Windows-only.")
return
try:
os.makedirs(self.config_dir, exist_ok=True)
dll_p = os.path.join(self.config_dir, DLL_NAME)
model_p = os.path.join(self.config_dir, MODEL_NAME)
if not os.path.exists(dll_p) or not os.path.exists(model_p):
msg = f"OneOCR init fail: DLL/Model not found in '{self.config_dir}'. See guide."
(self.logger or logging).warning(msg)
return
self.engine = OcrEngine(self.config_dir, self.logger)
self.available = True
if self.logger:
self.logger.info("OneOCR engine ready.")
except Exception as e:
if self.logger:
self.logger.error(
f"Failed to create OcrEngine: {e}", exc_info=self.debug_mode)
self.engine = None
self.available = False
def _ocr_blk_list(self, img: np.ndarray, blk_list: list[TextBlock], *args, **kwargs):
if not self.available:
return
im_h, im_w = img.shape[:2]
for i, blk in enumerate(blk_list):
x1, y1, x2, y2 = blk.xyxy
if 0 <= y1 < y2 <= im_h and 0 <= x1 < x2 <= im_w:
crop = img[y1:y2, x1:x2]
if crop.size == 0:
blk.text = ""
continue
try:
blk.text = self.ocr(crop, apply_postprocessing=True)
except Exception as e: # Log error from main ocr call
if self.logger:
self.logger.error(
f"OCR err block {i+1} {blk.xyxy}: {e}", exc_info=self.debug_mode)
blk.text = ""
else:
blk.text = "" # Invalid coords
def ocr_img(self, img: np.ndarray) -> str: return self.ocr(img,
apply_postprocessing=True) if self.available else ""
def ocr(self, img: np.ndarray, apply_postprocessing: bool = True) -> str:
if not self.available or self.engine is None or img is None or img.size == 0:
return ""
start_time = time.time()
original_h, original_w = img.shape[:2]
padded = False
if self.debug_mode and self.logger:
self.logger.debug(f"OCR start shape: {original_h}x{original_w}")
try:
img_to_process = img # Start with original image
if self.expand_small_blocks and (original_h < MIN_DIM_SIZE or original_w < MIN_DIM_SIZE):
pad_h_total = max(0, MIN_DIM_SIZE - original_h)
pad_w_total = max(0, MIN_DIM_SIZE - original_w)
pad_top = pad_h_total//2
pad_bottom = pad_h_total - pad_top
pad_left = pad_w_total//2
pad_right = pad_w_total - pad_left
# Determine padding color (white) based on channels
if len(img.shape) == 2:
color = 255 # Grayscale
elif img.shape[2] == 3:
color = (255, 255, 255) # BGR
else:
color = (255, 255, 255, 255) # BGRA (or assume if > 3)
img_to_process = cv2.copyMakeBorder(
img, pad_top, pad_bottom, pad_left, pad_right, cv2.BORDER_CONSTANT, value=color)
padded = True
if self.debug_mode and self.logger:
self.logger.debug(
f"Padded img from {original_h}x{original_w} to {img_to_process.shape[:2]} (min={MIN_DIM_SIZE})")
# Convert potentially padded image to RGB for PIL
if len(img_to_process.shape) == 2:
img_rgb = cv2.cvtColor(img_to_process, cv2.COLOR_GRAY2RGB)
elif img_to_process.shape[2] == 3:
img_rgb = img_to_process
elif img_to_process.shape[2] == 4:
img_rgb = cv2.cvtColor(img_to_process, cv2.COLOR_RGBA2RGB)
else:
raise ValueError(
f"Unsupported channels: {img_to_process.shape[2]}")
pil_image = PilImage.fromarray(img_rgb).convert("RGBA")
# This might raise RuntimeError on failure code
result_dict = self.engine.recognize_pil(pil_image)
lines = [line["text"]
for line in result_dict.get("lines", []) if line.get("text")]
if self.reverse_line_order:
lines.reverse()
full_text = "\n".join(lines)
if apply_postprocessing:
full_text = self._apply_text_postprocessing(full_text)
if self.debug_mode and self.logger:
self.logger.debug(
f"OCR done {(time.time()-start_time):.3f}s, padded: {padded}, lines: {len(lines)}")
return full_text
except RuntimeError as e: # Catch errors from _check_dll_result
err_code_match = re.search(r"\(Native Code: (\d+)\)", str(e))
err_code = int(err_code_match.group(1)) if err_code_match else None
log_msg = f"Critical OCR error: {e}"
# Check if it's code 3 and padding *wasn't* applied to a small image
if err_code == 3 and not padded and (original_h < MIN_DIM_SIZE or original_w < MIN_DIM_SIZE):
log_msg += f" (Native Code 3 often means image too small. Try enabling 'expand_small_blocks' in params if disabled, or check if padding to {MIN_DIM_SIZE}px is sufficient)"
if self.logger:
self.logger.error(log_msg, exc_info=self.debug_mode)
return ""
except Exception as e: # Catch other errors (PIL conversion, etc.)
if self.logger:
self.logger.error(
f"Unexpected OCR error: {e}", exc_info=self.debug_mode)
return ""
def _apply_text_postprocessing(self, text: str) -> str:
if not text:
return ""
if self.newline_handling == "remove":
text = text.replace("\n", " ").replace("\r", "")
elif self.newline_handling == "preserve":
text = text.replace("\r\n", "\n").replace("\r", "\n")
text = self._apply_punctuation_and_spacing(text)
if self.no_uppercase:
text = self._apply_no_uppercase(text)
return text
def _apply_no_uppercase(self, text: str) -> str:
if not text:
return ""
return " ".join([s[0].upper()+s[1:].lower() for s in re.split(r"(?<=[.!?…])\s+", text) if s])
def _apply_punctuation_and_spacing(self, text: str) -> str:
if not text:
return ""
# Remove space before punct
text = re.sub(r"\s+([,.!?…;:])", r"\1", text)
text = re.sub(r"([,.!?…;:])(?=[^\s,.!?…;:])", r"\1 ",
text) # Add space after punct if missing
text = re.sub(r"\s{2,}", " ", text) # Collapse multiple spaces
return text.strip()
def updateParam(self, key: str, content):
super().updateParam(key, content)
if self.debug_mode and self.logger:
self.logger.debug(f"Param '{key}' updated in OCROneAPI.")
|