Delete process_local_image.py

process_local_image.py

@@ -1,104 +0,0 @@
-from transformers import LayoutLMv3Processor, LayoutLMv3ForTokenClassification
-import torch
-from PIL import Image
-import json
-import pytesseract
-
-# ===== Load pretrained LayoutLMv3 model and processor =====
-MODEL_ID = "microsoft/layoutlmv3-base"
-processor = LayoutLMv3Processor.from_pretrained(MODEL_ID, apply_ocr=False)
-model = LayoutLMv3ForTokenClassification.from_pretrained(MODEL_ID)
-
-# ===== Define label mapping (placeholders, adjust as needed) =====
-labels = [
-    "O",            # Outside any entity
-    "B-NAME", "I-NAME",
-    "B-DOB", "I-DOB",
-    "B-ADDRESS", "I-ADDRESS",
-    "B-LICENSE", "I-LICENSE"
-]
-
-def extract_driver_license_info(image_path):
-    print(f"🔄 Processing image: {image_path}")
-    
-    # Load image
-    image = Image.open(image_path).convert("RGB")
-    print(f"📏 Image size: {image.size[0]}x{image.size[1]} pixels")
-    
-    # Step 1: Extract text using Tesseract OCR
-    print("🔍 Extracting text with Tesseract OCR...")
-    full_text = pytesseract.image_to_string(image)
-    print(f"📄 Tesseract OCR extracted: {len(full_text)} characters")
-    print(f"📝 Raw OCR text:\n{full_text}")
-    
-    # Step 2: Get bounding boxes from Tesseract for LayoutLMv3
-    print("🤖 Processing with LayoutLMv3 for document understanding...")
-    
-    try:
-        # Get detailed OCR data with bounding boxes
-        ocr_data = pytesseract.image_to_data(image, output_type=pytesseract.Output.DICT)
-        
-        # Filter out empty words and get bounding boxes
-        words = []
-        boxes = []
-        img_width, img_height = image.size
-        
-        for i in range(len(ocr_data['text'])):
-            word = ocr_data['text'][i].strip()
-            if word and int(ocr_data['conf'][i]) > 30:  # Confidence threshold
-                words.append(word)
-                x, y, w, h = ocr_data['left'][i], ocr_data['top'][i], ocr_data['width'][i], ocr_data['height'][i]
-                # Normalize bounding boxes to 0-1000 range as required by LayoutLMv3
-                x_norm = int((x / img_width) * 1000)
-                y_norm = int((y / img_height) * 1000)
-                x2_norm = int(((x + w) / img_width) * 1000)
-                y2_norm = int(((y + h) / img_height) * 1000)
-                boxes.append([x_norm, y_norm, x2_norm, y2_norm])
-        
-        print(f"📝 Found {len(words)} words with bounding boxes")
-        
-        if len(words) > 0 and len(boxes) > 0:
-            # Process with LayoutLMv3 using Tesseract's bounding boxes
-            encoding = processor(image, words, boxes=boxes, return_tensors="pt")
-            
-            with torch.no_grad():
-                outputs = model(**encoding)
-                logits = outputs.logits
-                predictions = logits.argmax(-1).squeeze().tolist()
-            
-            print(f"📝 LayoutLMv3 processed {len(words)} words with bounding boxes")
-            print(f"📊 All predictions: {predictions}")
-            print(f"📊 All labels: {[labels[p] if p < len(labels) else f'UNKNOWN_{p}' for p in predictions]}")
-            print(f"📊 Word-prediction pairs: {list(zip(words, predictions))}")
-            
-            # Extract entities using LayoutLMv3 predictions
-            extracted = {"name": "", "dob": "", "address": "", "license_number": "", "expiry": ""}
-            current_label = None
-            
-            for word, pred_id in zip(words, predictions):
-                if pred_id < len(labels):
-                    label = labels[pred_id]
-                    if label.startswith("B-"):
-                        current_label = label[2:].lower()
-                        extracted[current_label] = word
-                    elif label.startswith("I-") and current_label:
-                        extracted[current_label] += " " + word
-                    else:
-                        current_label = None
-            
-            print(f"📊 LayoutLMv3 extracted: {extracted}")
-            
-        else:
-            print("⚠️ No valid words found by Tesseract")
-            extracted = {"name": "", "dob": "", "address": "", "license_number": "", "expiry": ""}
-            
-    except Exception as e:
-        print(f"⚠️ LayoutLMv3 processing failed: {str(e)}")
-        extracted = {"name": "", "dob": "", "address": "", "license_number": "", "expiry": ""}
-    
-    return extracted
-
-
-if __name__ == "__main__":
-    result = extract_driver_license_info("1.jpg")
-    print(json.dumps(result, indent=4))