Update app.py

app.py

@@ -9,6 +9,7 @@ import re
 from dataclasses import dataclass
 from typing import List, Dict, Tuple
 
+
 import cv2
 import numpy as np
 import torch
@@ -16,8 +17,10 @@ from paddleocr import TextDetection
 from easyocr import Reader
 from rapidfuzz import fuzz
 
+
 import gradio as gr
 
+
 # ============ CORE VALIDATORS (UNCHANGED) ============
 class VerhoeffValidator:
     d_table = [[0,1,2,3,4,5,6,7,8,9],[1,2,3,4,0,6,7,8,9,5],[2,3,4,0,1,7,8,9,5,6],[3,4,0,1,2,8,9,5,6,7],[4,0,1,2,3,9,5,6,7,8],[5,9,8,7,6,0,4,3,2,1],[6,5,9,8,7,1,0,4,3,2],[7,6,5,9,8,2,1,0,4,3],[8,7,6,5,9,3,2,1,0,4],[9,8,7,6,5,4,3,2,1,0]]
@@ -29,6 +32,7 @@ class VerhoeffValidator:
         for i,ch in enumerate(reversed(n)): c=cls.d_table[c][cls.p_table[i%8][int(ch)]]
         return c==0
 
+
 class PatternValidator:
     @staticmethod
     def find_aadhaar(t: str) -> List[str]:
@@ -38,6 +42,7 @@ class PatternValidator:
     def find_pan(t: str) -> List[str]:
         return list(set(re.findall(r'\b[A-Z]{3}[PCHFATBLJG][A-Z]\d{4}[A-Z]\b', t.upper())))
 
+
 class TextNormalizer:
     OCR_CORRECTIONS = {'O':'0','o':'0','l':'1','I':'1','Z':'2','z':'2','S':'5','G':'6','b':'6','T':'7','B':'8','g':'9','q':'9'}
     @staticmethod
@@ -52,6 +57,7 @@ class TextNormalizer:
             text = re.sub(r'\b[0-9OolIZzSGbTBgq]{4,}\b', fix, text)
         return re.sub(r'\s+',' ',re.sub(r'[^\w\s\u0900-\u097F.,/-]','',text)).strip()
 
+
 # ============ CONFIGURATION ============
 @dataclass
 class Config:
@@ -71,6 +77,7 @@ class Config:
                 "Ration_Card": ["ration","card","food","civil","supplies","apl","bpl","राशन","कार्ड","खाद्य","नागरी","पुरवठा"]
             }
 
+
 # ============ MAIN PIPELINE ============
 class DocumentOCRVerifier:
     def __init__(self, config: Config=None):
@@ -82,11 +89,13 @@ class DocumentOCRVerifier:
             self.detector = None
         self.reader = Reader(self.cfg.languages, gpu=torch.cuda.is_available())
 
+
     def _preprocess(self, img: np.ndarray) -> np.ndarray:
         img = self._resize(img)
         img = self._deskew(img)
         return self._enhance(img)
 
+
     def _resize(self, img: np.ndarray) -> np.ndarray:
         h,w = img.shape[:2]
         if max(h,w) > self.cfg.max_image_dim:
@@ -94,6 +103,7 @@ class DocumentOCRVerifier:
             img = cv2.resize(img, (int(w*scale), int(h*scale)), interpolation=cv2.INTER_AREA)
         return img
 
+
     def _deskew(self, img: np.ndarray) -> np.ndarray:
         gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
         _, binary = cv2.threshold(gray, 0, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)
@@ -109,6 +119,7 @@ class DocumentOCRVerifier:
                 img = cv2.warpAffine(img, M, (w,h), borderValue=(255,255,255))
         return img
 
+
     def _enhance(self, img: np.ndarray) -> np.ndarray:
         denoised = cv2.fastNlMeansDenoisingColored(img, None, 10, 10, 7, 21)
         lab = cv2.cvtColor(denoised, cv2.COLOR_BGR2LAB)
@@ -118,10 +129,12 @@ class DocumentOCRVerifier:
         kernel = np.array([[0,-1,0],[-1,5,-1],[0,-1,0]])
         return cv2.addWeighted(cv2.filter2D(enhanced, -1, kernel), 0.6, enhanced, 0.4, 0)
 
+
     def _extract_keywords(self, text: str) -> List[str]:
         if not text: return []
         return [t for t in re.split(r'\s+', text.strip()) if t]
 
+
     def _classify(self, text: str) -> Tuple[str, float, List[str]]:
         norm_text = TextNormalizer.normalize(text, aggressive=True)
         scores = {}
@@ -140,20 +153,24 @@ class DocumentOCRVerifier:
             scores[doc_type] = {"score": score, "matched": matched}
         winner = max(scores.items(), key=lambda x: x[1]["score"])
         if winner[1]["score"] >= self.cfg.min_keywords:
-            conf = 0.95 if winner[1]["score"] == 100 else min(0.90, len(winner[1]["matched")]/len(self.cfg.doc_keywords[winner[0]]) + 0.3) if len(self.cfg.doc_keywords[winner[0]])>0 else 0.0
+            conf = 0.95 if winner[1]["score"] == 100 else min(0.90, len(winner[1]["matched"])/len(self.cfg.doc_keywords[winner[0]]) + 0.3)
             return winner[0], conf, winner[1]["matched"]
         return "UNCLASSIFIED", 0.0, []
 
+
     def verify(self, image_path: str, user_keywords: List[str]) -> Dict:
         img = cv2.imread(image_path)
         if img is None: return {"error": "Image not found", "imagePath": image_path}
 
+
         img = self._preprocess(img)
 
+
         # Region-based OCR with word-level granularity
         ocr_keywords = []
         all_text = ""
 
+
         if self.detector:
             try:
                 regions = self.detector.predict(input=image_path, batch_size=1)
@@ -162,6 +179,7 @@ class DocumentOCRVerifier:
         else:
             regions = []
 
+
         # If detector provided regions, use them; otherwise fallback to whole-image read
         if regions:
             for res in regions:
@@ -183,15 +201,20 @@ class DocumentOCRVerifier:
                     ocr_keywords.extend(self._extract_keywords(t))
                     all_text += " " + t
 
+
         # Classification
         doc_type, accuracy, matched_keywords = self._classify(all_text)
 
+
         # Verification - match against combined text for phrase support
+        # Preserve raw input keywords (split externally) but perform exact matching on the combined OCR text without further altering user's internal spacing
         raw_input_keywords = user_keywords
+        # Do minimal trimming for matching (only strip outer whitespace)
         minimal_norm_user_keywords = [kw.strip() for kw in raw_input_keywords if kw is not None]
         exact_matches = list(set([kw for kw in minimal_norm_user_keywords if kw.lower() in all_text.lower()]))
         status = "verified" if exact_matches else "not_verified"
 
+
         return {
             "documentType": doc_type,
             "documentTypeAccuracy": round(accuracy, 4),
@@ -203,8 +226,10 @@ class DocumentOCRVerifier:
             "imagePath": image_path
         }
 
+
 # ============ APP ============
 
+
 verifier = DocumentOCRVerifier()
 
 
@@ -235,7 +260,7 @@ def save_upload_to_tmp(uploaded_file) -> str:
             with open(uploaded_file, "rb") as src, open(out_path, "wb") as dst:
                 dst.write(src.read())
         except Exception:
-            # last resort: try to read as PIL image (if provided)
+            # last resort: try to read as numpy array (if provided)
             try:
                 import PIL.Image as Image
                 im = Image.open(uploaded_file).convert("RGB")
@@ -247,94 +272,147 @@ def save_upload_to_tmp(uploaded_file) -> str:
 
 def run_ocr(image, keywords_raw: str):
     """
-    image: uploaded file path or bytes (Gradio File component)
+    image: uploaded file path or bytes (Gradio Image component with type='file' or 'numpy')
     keywords_raw: raw string entered by user. Split by comma EXACTLY to form keywords. Preserve internal spacing.
-    Returns: image preview (path), HTML summary, parsed JSON (dict) for gr.JSON, raw JSON string
     """
-    # Parse user keywords preserving internal spaces
+    if image is None:
+        return None, "<div style='color: red; padding: 20px;'>⚠️ Please upload an image first!</div>", ""
+    
+    # Split user keywords by comma only; do not auto-trim internal spaces (only strip ends)
     if keywords_raw is None:
         user_keywords = []
     else:
+        # Split on commas. Keep empty tokens if user left them intentionally.
         user_keywords = [s if s is not None else "" for s in re.split(r',', keywords_raw)]
+        # strip only leading/trailing newline and tabs, but preserve internal spacing and common spaces
         user_keywords = [s.rstrip("\n\r\t ").lstrip("\n\r\t ") for s in user_keywords]
 
+
+    # Save file to /tmp and call verifier
     image_path = save_upload_to_tmp(image)
     result = verifier.verify(image_path=image_path, user_keywords=user_keywords)
-
-    # build a simple card-style HTML summary with color scheme for verification status
-    status = result.get("verificationStatus", "not_verified")
-    doc_type = result.get("documentType", "UNCLASSIFIED")
-    accuracy = result.get("documentTypeAccuracy", 0.0)
-    input_keys = result.get("inputUserKeywords", [])
-
-    status_color = "#1a7f37" if status == "verified" else "#b22222"
+    
+    # Extract fields for card display
+    doc_type = result.get("documentType", "N/A")
+    doc_accuracy = result.get("documentTypeAccuracy", 0.0)
+    input_keywords = result.get("inputUserKeywords", [])
+    verification_status = result.get("verificationStatus", "not_verified")
+    
+    # Format accuracy as percentage
+    accuracy_text = f"{doc_accuracy * 100:.2f}%"
+    
+    # Format keywords as comma-separated string
+    keywords_text = ", ".join([f'"{kw}"' for kw in input_keywords]) if input_keywords else "None provided"
+    
+    # Color-coded status
+    if verification_status == "verified":
+        status_html = '<span style="color: #16a34a; font-weight: bold; font-size: 22px;">✓ VERIFIED</span>'
+        status_bg = "#dcfce7"
+        status_border = "#16a34a"
+    else:
+        status_html = '<span style="color: #dc2626; font-weight: bold; font-size: 22px;">✗ NOT VERIFIED</span>'
+        status_bg = "#fee2e2"
+        status_border = "#dc2626"
+    
+    # Create HTML card with improved styling
     card_html = f"""
-    <div style='border-radius:8px;padding:14px;box-shadow:0 4px 12px rgba(0,0,0,0.08);max-width:640px;font-family:system-ui, -apple-system, Segoe UI, Roboto, "Helvetica Neue", Arial;'>
-      <div style='display:flex;align-items:center;justify-content:space-between;margin-bottom:8px;'>
-        <div style='font-size:16px;font-weight:600'>Document Summary</div>
-        <div style='padding:6px 10px;border-radius:14px;background:{status_color};color:white;font-weight:700'>{status.upper()}</div>
-      </div>
-      <div style='display:grid;grid-template-columns:1fr 1fr;gap:8px;'>
-        <div style='background:#fafafa;padding:10px;border-radius:6px;'>
-          <div style='font-size:12px;color:#555'>Document Type</div>
-          <div style='font-size:15px;font-weight:600'>{doc_type}</div>
-        </div>
-        <div style='background:#fafafa;padding:10px;border-radius:6px;'>
-          <div style='font-size:12px;color:#555'>Document Accuracy</div>
-          <div style='font-size:15px;font-weight:600'>{accuracy*100:.2f}%</div>
+    <div style="border: 2px solid #e5e7eb; border-radius: 16px; padding: 28px; background: linear-gradient(135deg, #ffffff 0%, #f9fafb 100%); box-shadow: 0 10px 25px rgba(0,0,0,0.1); font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif;">
+        <div style="display: flex; align-items: center; margin-bottom: 24px; border-bottom: 3px solid #3b82f6; padding-bottom: 16px;">
+            <span style="font-size: 32px; margin-right: 12px;">📄</span>
+            <h2 style="margin: 0; color: #1f2937; font-size: 26px; font-weight: 700;">Document Verification Results</h2>
         </div>
-        <div style='grid-column:1 / -1;background:#fff8e6;padding:10px;border-radius:6px;'>
-          <div style='font-size:12px;color:#555'>Input Keywords</div>
-          <div style='font-size:14px;font-weight:600'>{', '.join(input_keys) if input_keys else '—'}</div>
+        
+        <div style="display: grid; gap: 16px;">
+            <div style="background: linear-gradient(135deg, #dbeafe 0%, #eff6ff 100%); padding: 20px; border-radius: 12px; border-left: 5px solid #3b82f6; box-shadow: 0 2px 8px rgba(59,130,246,0.2);">
+                <div style="color: #1e40af; font-size: 13px; font-weight: 600; text-transform: uppercase; letter-spacing: 0.5px; margin-bottom: 8px;">📋 Document Type</div>
+                <div style="font-size: 24px; color: #1f2937; font-weight: 700;">{doc_type}</div>
+            </div>
+            
+            <div style="background: linear-gradient(135deg, #d1fae5 0%, #ecfdf5 100%); padding: 20px; border-radius: 12px; border-left: 5px solid #10b981; box-shadow: 0 2px 8px rgba(16,185,129,0.2);">
+                <div style="color: #065f46; font-size: 13px; font-weight: 600; text-transform: uppercase; letter-spacing: 0.5px; margin-bottom: 8px;">🎯 Type Detection Accuracy</div>
+                <div style="font-size: 24px; color: #1f2937; font-weight: 700;">{accuracy_text}</div>
+            </div>
+            
+            <div style="background: linear-gradient(135deg, #fef3c7 0%, #fefce8 100%); padding: 20px; border-radius: 12px; border-left: 5px solid #f59e0b; box-shadow: 0 2px 8px rgba(245,158,11,0.2);">
+                <div style="color: #92400e; font-size: 13px; font-weight: 600; text-transform: uppercase; letter-spacing: 0.5px; margin-bottom: 8px;">🔑 Input Keywords</div>
+                <div style="font-size: 17px; color: #451a03; font-weight: 500; line-height: 1.6;">{keywords_text}</div>
+            </div>
+            
+            <div style="background: linear-gradient(135deg, {status_bg} 0%, {status_bg}dd 100%); padding: 24px; border-radius: 12px; border: 3px solid {status_border}; text-align: center; box-shadow: 0 4px 12px rgba(0,0,0,0.15);">
+                <div style="color: #374151; font-size: 13px; font-weight: 600; text-transform: uppercase; letter-spacing: 0.5px; margin-bottom: 12px;">🔍 Verification Status</div>
+                <div style="margin-top: 8px;">{status_html}</div>
+            </div>
         </div>
-      </div>
     </div>
     """
-
-    raw_json_str = json.dumps(result, indent=2, ensure_ascii=False)
-
-    # return values: image preview path (Gradio Image accepts file path), summary HTML, parsed JSON, raw JSON string
-    return image_path, card_html, result, raw_json_str
-
-
-with gr.Blocks(title="Document OCR Verifier") as demo:
-    gr.Markdown("Upload an image and provide comma-separated keywords (preserve spacing inside keywords).")
+    
+    # Return JSON string exactly as produced
+    json_output = json.dumps(result, indent=2, ensure_ascii=False)
+    
+    return image_path, card_html, json_output
+
+
+# Custom CSS for better styling
+custom_css = """
+.gradio-container {
+    font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif !important;
+}
+
+.gr-button-primary {
+    background: linear-gradient(90deg, #3b82f6 0%, #2563eb 100%) !important;
+    border: none !important;
+    font-weight: 600 !important;
+    font-size: 16px !important;
+    padding: 12px 24px !important;
+    transition: all 0.3s ease !important;
+}
+
+.gr-button-primary:hover {
+    transform: translateY(-2px) !important;
+    box-shadow: 0 8px 16px rgba(59, 130, 246, 0.3) !important;
+}
+
+.gr-box {
+    border-radius: 12px !important;
+}
+"""
+
+with gr.Blocks(title="Document OCR Verifier", css=custom_css) as demo:
+    gr.Markdown("""
+    # 🔍 Document OCR Verifier
+    ### Upload a document image and provide comma-separated keywords to verify the document authenticity.
+    """)
+    
     with gr.Row():
-        img_in = gr.File(label="Image (JPEG/PNG). The file will be stored in /tmp.")
-        kws = gr.Textbox(label="User keywords (comma-separated). Example: ROHIT, KUMAR, SINGH")
-    run_btn = gr.Button("Run OCR & Verify")
-
+        with gr.Column(scale=1):
+            img_in = gr.File(label="📤 Upload Document Image (JPEG/PNG)")
+            kws = gr.Textbox(
+                label="🔑 Verification Keywords (comma-separated)", 
+                placeholder="Example: ROHIT, KUMAR, SINGH, Date of Birth",
+                lines=3
+            )
+            run_btn = gr.Button("🚀 Run OCR & Verify", variant="primary", size="lg")
+        
+        with gr.Column(scale=1):
+            img_out = gr.Image(label="📸 Uploaded Document", type="filepath", height=400)
+    
     with gr.Row():
-        with gr.Column():
-            card = gr.HTML("""
-            <div style='border:1px solid #ccc;padding:16px;border-radius:12px;width:100%;'>
-                <h3>Document Summary</h3>
-                <p><strong>Document Type:</strong> <span id='doc_type'></span></p>
-                <p><strong>Accuracy:</strong> <span id='doc_acc'></span></p>
-                <p><strong>User Keywords:</strong> <span id='user_kws'></span></p>
-                <p><strong>Status:</strong> <span id='verif_status' style='font-weight:bold;'></span></p>
-            </div>
-            """)
-        out = gr.Textbox(label="Result (raw JSON)", lines=20)
-
-    def update_card(json_str):
-        try:
-            data = json.loads(json_str)
-            status_color = "green" if data.get("verificationStatus") == "verified" else "red"
-            html = f"""
-            <div style='border:1px solid #ccc;padding:16px;border-radius:12px;width:100%;'>
-                <h3>Document Summary</h3>
-                <p><strong>Document Type:</strong> {data.get('documentType')}</p>
-                <p><strong>Accuracy:</strong> {data.get('documentTypeAccuracy')}</p>
-                <p><strong>User Keywords:</strong> {', '.join(data.get('inputUserKeywords', []))}</p>
-                <p><strong>Status:</strong> <span style='color:{status_color};font-weight:bold;'>{data.get('verificationStatus')}</span></p>
-            </div>
-            """
-            return html
-        except:
-            return "<div>Invalid JSON</div>"
-
-    run_btn.click(fn=run_ocr, inputs=[img_in, kws], outputs=[out])
-    out.change(fn=update_card, inputs=[out], outputs=[card])
-
-demo.launch(server_name="0.0.0.0", server_port=7860)(server_name="0.0.0.0", server_port=7860)
+        card_out = gr.HTML(label="📊 Verification Summary")
+    
+    with gr.Row():
+        json_out = gr.Textbox(label="📋 Complete JSON Response", lines=18, max_lines=25)
+    
+    run_btn.click(
+        fn=run_ocr, 
+        inputs=[img_in, kws], 
+        outputs=[img_out, card_out, json_out]
+    )
+    
+    gr.Markdown("""
+    ---
+    **Note:** The document will be stored in `/tmp/ocr_app/` directory. Supported formats: JPEG, PNG, JPG.
+    """)
+
+
+if __name__ == "__main__":
+    demo.launch(server_name="0.0.0.0", server_port=7860)