Update app.py

app.py

@@ -57,22 +57,64 @@ def extract_text(image_path):
 
 def detect_document(text):
 
+    # CLEAN TEXT
+    text = text.lower().strip()
+
+    # REMOVE EXTRA SYMBOLS
+    cleaned_text = re.sub(
+        r'[^a-zA-Z0-9\\s-]',
+        ' ',
+        text
+    )
+
+    # SPLIT WORDS
+    words = cleaned_text.split()
+
+    # =========================
+    # REJECT RANDOM OCR GARBAGE
+    # =========================
+
+    garbage_patterns = [
+        r'^[a-z0-9]{4,8}$'
+    ]
+
+    for pattern in garbage_patterns:
+
+        for word in words:
+
+            if re.match(pattern, word):
+
+                if len(words) <= 2:
+                    return {
+                        "document_type": "unknown",
+                        "confidence": 5,
+                        "matched_keywords": [word],
+                        "reason": (
+                            "OCR detected unreadable or "
+                            "meaningless text."
+                        )
+                    }
+
     # =========================
     # NIN
     # =========================
 
     nin_keywords = [
         "national identification number",
-        "nin"
+        "national identity",
+        "nin",
+        "nimc"
     ]
 
     matched_keywords = []
 
     for keyword in nin_keywords:
-        if keyword in text:
+
+        if keyword in cleaned_text:
             matched_keywords.append(keyword)
 
     if len(matched_keywords) > 0:
+
         return {
             "document_type": "nin",
             "confidence": 95,
@@ -85,16 +127,19 @@ def detect_document(text):
 
     passport_keywords = [
         "passport",
-        "federal republic of nigeria"
+        "federal republic of nigeria",
+        "nigeria passport"
     ]
 
     matched_keywords = []
 
     for keyword in passport_keywords:
-        if keyword in text:
+
+        if keyword in cleaned_text:
             matched_keywords.append(keyword)
 
     if len(matched_keywords) > 0:
+
         return {
             "document_type": "passport",
             "confidence": 94,
@@ -107,16 +152,21 @@ def detect_document(text):
 
     license_keywords = [
         "driver",
-        "license"
+        "license",
+        "drivers licence",
+        "driver licence",
+        "frsc"
     ]
 
     matched_keywords = []
 
     for keyword in license_keywords:
-        if keyword in text:
+
+        if keyword in cleaned_text:
             matched_keywords.append(keyword)
 
     if len(matched_keywords) >= 2:
+
         return {
             "document_type": "drivers_license",
             "confidence": 92,
@@ -129,16 +179,20 @@ def detect_document(text):
 
     voter_keywords = [
         "voter",
-        "inec"
+        "inec",
+        "permanent voter",
+        "polling unit"
     ]
 
     matched_keywords = []
 
     for keyword in voter_keywords:
-        if keyword in text:
+
+        if keyword in cleaned_text:
             matched_keywords.append(keyword)
 
     if len(matched_keywords) > 0:
+
         return {
             "document_type": "voters_card",
             "confidence": 90,
@@ -146,28 +200,61 @@ def detect_document(text):
         }
 
     # =========================
-    # UTILITY BILL
+    # ELECTRICITY COMPANIES
     # =========================
 
-    utility_keywords = [
+    electricity_keywords = [
+
+        # General
         "electricity",
-        "water bill",
+        "electric bill",
+        "power bill",
+        "meter number",
+
+        # Nigerian DISCOs
+        "ibedc",
+        "ibadan electricity",
+        "ikedc",
         "ikeja electric",
+        "ekedc",
         "eko electric",
+        "aedc",
         "abuja electricity",
-        "aedc"
+        "eedc",
+        "enugu electricity",
+        "bedc",
+        "benin electricity",
+        "jed",
+        "jos electricity",
+        "kedco",
+        "kano electricity",
+        "kaedco",
+        "kaduna electric",
+        "phed",
+        "port harcourt electricity",
+        "yedc",
+        "yola electricity",
+
+        # Common Nigerian utility terms
+        "prepaid",
+        "postpaid",
+        "disco",
+        "energy charge",
+        "tariff"
     ]
 
     matched_keywords = []
 
-    for keyword in utility_keywords:
-        if keyword in text:
+    for keyword in electricity_keywords:
+
+        if keyword in cleaned_text:
             matched_keywords.append(keyword)
 
     if len(matched_keywords) > 0:
+
         return {
             "document_type": "utility_bill",
-            "confidence": 88,
+            "confidence": 90,
             "matched_keywords": matched_keywords
         }
 
@@ -176,21 +263,39 @@ def detect_document(text):
     # =========================
 
     bank_keywords = [
+
         "account statement",
+        "statement of account",
         "transaction",
         "balance",
         "account number",
         "credit",
-        "debit"
+        "debit",
+        "withdrawal",
+        "deposit",
+
+        # Nigerian Banks
+        "access bank",
+        "gtbank",
+        "uba",
+        "zenith bank",
+        "first bank",
+        "opay",
+        "moniepoint",
+        "kuda",
+        "fcmb",
+        "sterling bank"
     ]
 
     matched_keywords = []
 
     for keyword in bank_keywords:
-        if keyword in text:
+
+        if keyword in cleaned_text:
             matched_keywords.append(keyword)
 
     if len(matched_keywords) > 0:
+
         return {
             "document_type": "bank_statement",
             "confidence": 91,
@@ -202,20 +307,25 @@ def detect_document(text):
     # =========================
 
     tenancy_keywords = [
+
         "tenancy agreement",
         "landlord",
         "tenant",
         "rent",
-        "property"
+        "property",
+        "lease agreement",
+        "rental agreement"
     ]
 
     matched_keywords = []
 
     for keyword in tenancy_keywords:
-        if keyword in text:
+
+        if keyword in cleaned_text:
             matched_keywords.append(keyword)
 
     if len(matched_keywords) > 0:
+
         return {
             "document_type": "tenancy_agreement",
             "confidence": 89,
@@ -223,38 +333,142 @@ def detect_document(text):
         }
 
     # =========================
-    # VEHICLE PLATE NUMBER
+    # VEHICLE KEYWORDS
+    # =========================
+
+    vehicle_keywords = [
+
+        "toyota",
+        "honda",
+        "lexus",
+        "benz",
+        "mercedes",
+        "ford",
+        "jeep",
+        "hyundai",
+        "kia",
+        "nissan",
+        "camry",
+        "corolla",
+        "rav4",
+        "pilot",
+        "highlander",
+        "vehicle",
+        "plate number"
+    ]
+
+    matched_keywords = []
+
+    for keyword in vehicle_keywords:
+
+        if keyword in cleaned_text:
+            matched_keywords.append(keyword)
+
+    # =========================
+    # NIGERIAN STATES
+    # =========================
+
+    nigeria_states = [
+
+        "lagos",
+        "abuja",
+        "kano",
+        "kaduna",
+        "oyo",
+        "ogun",
+        "ondo",
+        "osun",
+        "kwara",
+        "imo",
+        "anambra",
+        "enugu",
+        "rivers",
+        "delta",
+        "edo",
+        "cross river",
+        "akwa ibom",
+        "bayelsa",
+        "plateau",
+        "benue",
+        "kogi",
+        "ekiti",
+        "niger",
+        "zamfara",
+        "sokoto",
+        "katsina",
+        "borno",
+        "yobe",
+        "adamawa",
+        "taraba",
+        "gombe",
+        "bauchi",
+        "jigawa",
+        "nasarawa",
+        "kebbi",
+        "ebonyi"
+    ]
+
+    state_matches = []
+
+    for state in nigeria_states:
+
+        if state in cleaned_text:
+            state_matches.append(state)
+
+    # =========================
+    # NIGERIAN PLATE PATTERNS
     # =========================
 
     plate_patterns = [
 
-    # Standard Nigerian plates
-    r"[A-Z]{2,3}\s?\d{2,3}[A-Z]{1,2}",
+        r"[A-Z]{3}-?\\d{3}[A-Z]{2}",
+        r"[A-Z]{2}\\d{3}[A-Z]{3}",
+        r"[A-Z]{3}\\s\\d{3}\\s[A-Z]{2}"
+    ]
+
+    detected_plate = None
 
-    # OCR damaged versions
-    r"[A-Z]{1,3}\d{2,3}[A-Z]{1,3}",
+    for pattern in plate_patterns:
 
-    # Hyphen formats
-    r"[A-Z]{2,3}-?\d{2,3}-?[A-Z]{1,2}"
-]
+        plate_match = re.search(
+            pattern,
+            cleaned_text.upper()
+        )
 
-for pattern in plate_patterns:
+        if plate_match:
 
-    plate_match = re.search(
-        pattern,
-        text.upper()
-    )
+            detected_plate = plate_match.group()
+
+            break
+
+    # =========================
+    # VEHICLE DETECTION
+    # =========================
 
-    if plate_match:
+    if detected_plate:
 
         return {
             "document_type": "vehicle_plate",
-            "confidence": 85,
-            "matched_keywords": [plate_match.group()]
+            "confidence": 97,
+            "matched_keywords": [
+                detected_plate
+            ] + state_matches
         }
 
-    return None
+    # VEHICLE WITHOUT CLEAR PLATE
+    if len(matched_keywords) > 0:
 
+        return {
+            "document_type": "vehicle_image",
+            "confidence": 75,
+            "matched_keywords": matched_keywords
+        }
+
+    # =========================
+    # UNKNOWN DOCUMENT
+    # =========================
+
+    return None
 
 # =========================
 # HOME ROUTE