95 lines
3.5 KiB
Python

#////////// PROJECT FILE SEPARATION LINE ////////// CODE AFTER THIS LINE ARE FROM: <ProjectRoot>/Data/Python_API_Endpoints/ocr_engines.py
import os
import io
import base64
import torch
import pytesseract
import easyocr
import numpy as np
import platform
from PIL import Image
# ---------------------------------------------------------------------
# Configure cross-platform Tesseract path
# ---------------------------------------------------------------------
BASE_DIR = os.path.dirname(os.path.abspath(__file__))
SYSTEM = platform.system()
if SYSTEM == "Windows":
TESSERACT_FOLDER = os.path.join(BASE_DIR, "Tesseract-OCR")
TESSERACT_EXE = os.path.join(TESSERACT_FOLDER, "tesseract.exe")
TESSDATA_DIR = os.path.join(TESSERACT_FOLDER, "tessdata")
if not os.path.isfile(TESSERACT_EXE):
raise EnvironmentError("Missing tesseract.exe in /Tesseract-OCR. Ensure the full folder is copied.")
pytesseract.pytesseract.tesseract_cmd = TESSERACT_EXE
os.environ["TESSDATA_PREFIX"] = TESSDATA_DIR
else:
# Assume Linux/macOS with system-installed Tesseract
pytesseract.pytesseract.tesseract_cmd = "tesseract"
# ---------------------------------------------------------------------
# EasyOCR Global Instances
# ---------------------------------------------------------------------
easyocr_reader_cpu = None
easyocr_reader_gpu = None
def initialize_ocr_engines():
global easyocr_reader_cpu, easyocr_reader_gpu
if easyocr_reader_cpu is None:
easyocr_reader_cpu = easyocr.Reader(['en'], gpu=False)
if easyocr_reader_gpu is None:
easyocr_reader_gpu = easyocr.Reader(['en'], gpu=torch.cuda.is_available())
# ---------------------------------------------------------------------
# Main OCR Handler
# ---------------------------------------------------------------------
def run_ocr_on_base64(image_b64: str, engine: str = "tesseract", backend: str = "cpu") -> list[str]:
if not image_b64:
raise ValueError("No base64 image data provided.")
try:
raw_bytes = base64.b64decode(image_b64)
image = Image.open(io.BytesIO(raw_bytes)).convert("RGB")
except Exception as e:
raise ValueError(f"Invalid base64 image input: {e}")
engine = engine.lower().strip()
backend = backend.lower().strip()
if engine in ["tesseract", "tesseractocr"]:
try:
text = pytesseract.image_to_string(image, config="--psm 6 --oem 1")
except pytesseract.TesseractNotFoundError:
raise RuntimeError("Tesseract binary not found or not available on this platform.")
elif engine == "easyocr":
initialize_ocr_engines()
reader = easyocr_reader_gpu if backend == "gpu" else easyocr_reader_cpu
result = reader.readtext(np.array(image), detail=1)
# Group by Y position (line-aware sorting)
result = sorted(result, key=lambda r: r[0][0][1])
lines = []
current_line = []
last_y = None
line_threshold = 10
for (bbox, text, _) in result:
y = bbox[0][1]
if last_y is None or abs(y - last_y) < line_threshold:
current_line.append(text)
else:
lines.append(" ".join(current_line))
current_line = [text]
last_y = y
if current_line:
lines.append(" ".join(current_line))
text = "\n".join(lines)
else:
raise ValueError(f"OCR engine '{engine}' not recognized.")
return [line.strip() for line in text.splitlines() if line.strip()]