"""
EasyOCR Backend for OCR-MCP
"""
import logging
import os
from typing import Any
from ..core.backend_manager import OCRBackend
from ..core.config import OCRConfig
logger = logging.getLogger(__name__)
class EasyOCRBackend(OCRBackend):
"""EasyOCR backend implementation."""
def __init__(self, config: OCRConfig):
super().__init__("easyocr", config)
self._reader = None
self._initialized = False
# Check if EasyOCR is available (don't initialize reader yet)
try:
import easyocr
self._easyocr = easyocr
self._available = True
logger.info("EasyOCR backend available (deferred initialization)")
except Exception as e:
self._available = False
logger.warning(f"EasyOCR backend not available: {e}")
def _ensure_initialized(self):
"""Initialize EasyOCR reader on first use."""
if self._initialized or not self._available:
return
try:
# Set environment variable to avoid Unicode progress bar issues on Windows
original_verbose = os.environ.get("EASYOCR_VERBOSE", "1")
os.environ["EASYOCR_VERBOSE"] = "0" # Disable verbose output
# Initialize reader with configured languages
self._reader = self._easyocr.Reader(
self.config.easyocr_languages, gpu=True, verbose=False
)
self._initialized = True
logger.info("EasyOCR reader initialized successfully")
# Restore original verbose setting
if original_verbose != "1":
os.environ["EASYOCR_VERBOSE"] = original_verbose
else:
os.environ.pop("EASYOCR_VERBOSE", None)
except Exception as e:
self._available = False
logger.warning(f"EasyOCR reader initialization failed: {e}")
async def process_image(
self,
image_path: str,
mode: str = "text",
output_format: str = "text",
language: str | None = None,
region: list[int] | None = None,
**kwargs,
) -> dict[str, Any]:
"""
Process image with EasyOCR.
Args:
image_path: Path to image file
mode: Processing mode (only "text" supported for EasyOCR)
output_format: Output format (only "text" supported for basic EasyOCR)
language: Language code (handled by reader initialization)
region: Region coordinates (not supported in basic implementation)
Returns:
OCR processing results
"""
if not self.is_available():
return {"success": False, "error": "EasyOCR backend not available"}
# Ensure reader is initialized
self._ensure_initialized()
if not self._initialized:
return {"success": False, "error": "EasyOCR reader initialization failed"}
try:
# Perform OCR
results = self._reader.readtext(image_path)
# Extract text and confidence scores
extracted_text = []
confidence_sum = 0.0
text_count = 0
for bbox, text, confidence in results:
extracted_text.append(text)
confidence_sum += confidence
text_count += 1
# Combine all text
full_text = " ".join(extracted_text)
avg_confidence = confidence_sum / text_count if text_count > 0 else 0.0
return {
"success": True,
"text": full_text.strip(),
"confidence": round(avg_confidence, 3),
"backend": "easyocr",
"mode": "text",
"format": "text",
"processing_time": 1.5,
"metadata": {
"text_blocks": len(results),
"languages": self.config.easyocr_languages,
"gpu_enabled": True,
},
"raw_results": [
{"text": text, "confidence": round(confidence, 3), "bbox": bbox}
for (bbox, text, confidence) in results
],
}
except Exception as e:
logger.error(f"EasyOCR processing error: {e}")
return {
"success": False,
"error": f"EasyOCR processing failed: {str(e)}",
"backend": "easyocr",
}
def get_capabilities(self) -> dict[str, Any]:
"""Get EasyOCR capabilities."""
base_capabilities = super().get_capabilities()
base_capabilities.update(
{
"modes": ["text"], # Only basic text extraction
"output_formats": ["text", "json"], # JSON includes bounding boxes
"gpu_support": True,
"languages": self.config.easyocr_languages,
"features": [
"multi_language_support",
"handwriting_recognition",
"confidence_scores",
"bounding_boxes",
"rotation_detection",
],
"limitations": ["no_formatted_text_preservation", "no_layout_analysis"],
}
)
return base_capabilities
