Upload 6 files

app.py

@@ -11,19 +11,19 @@ import io
 import tempfile
 import torchvision
 
-
 class LicensePlateProcessor:
     def __init__(self):
         # Load models for plate detection
-        self.yolo_detector = YOLO('models/best.pt')  # For plate detection
-        self.char_reader = YOLO('models/read_char.pt')  # For character reading
+        self.yolo_detector = YOLO('detect_plate.pt')  # For license plate detection
+        self.province_detector = YOLO('best.pt')  # For province detection
+        self.char_reader = YOLO('read_char.pt')  # For character reading
         
         # Load TrOCR for province detection
         self.processor_plate = TrOCRProcessor.from_pretrained('openthaigpt/thai-trocr')
         self.model_plate = VisionEncoderDecoderModel.from_pretrained('openthaigpt/thai-trocr')
         
         # Load character mapping from yaml
-        with open('config/data.yaml', 'r', encoding='utf-8') as f:
+        with open('data.yaml', 'r', encoding='utf-8') as f:
             data_config = yaml.safe_load(f)
             self.char_mapping = data_config.get('char_mapping', {})
             self.names = data_config['names']
@@ -100,14 +100,16 @@ class LicensePlateProcessor:
                 return None
 
             # Detect license plate location
-            results = self.yolo_detector(image)
+            plate_results = self.yolo_detector(image)
+            province_results = self.province_detector(image)
             
             data = {"plate_number": "", "province": "", "raw_province": ""}
             
             # Save visualization
             output_image = image.copy()
             
-            for result in results:
+            # Process license plate detections
+            for result in plate_results:
                 for box in result.boxes:
                     confidence = float(box.conf)
                     if confidence < self.CONF_THRESHOLD:
@@ -116,29 +118,39 @@ class LicensePlateProcessor:
                     x1, y1, x2, y2 = map(int, box.xyxy.flatten())
                     cropped_image = image[y1:y2, x1:x2]
 
-                    # Draw rectangle on output image
-                    color = (0, 255, 0) if int(box.cls.item()) == 0 else (255, 0, 0)
-                    cv2.rectangle(output_image, (x1, y1), (x2, y2), color, 2)
+                    # Draw rectangle on output image (green for plate)
+                    cv2.rectangle(output_image, (x1, y1), (x2, y2), (0, 255, 0), 2)
+                    
+                    # Read characters using YOLO character reader
+                    data["plate_number"] = self.read_plate_characters(cropped_image)
 
-                    if int(box.cls.item()) == 0:  # License plate number
-                        # Read characters using YOLO character reader
-                        data["plate_number"] = self.read_plate_characters(cropped_image)
-                        
-                    elif int(box.cls.item()) == 1:  # Province
-                        # Process province using TrOCR
-                        cropped_image_gray = cv2.cvtColor(cropped_image, cv2.COLOR_BGR2GRAY)
-                        equalized_image = cv2.equalizeHist(cropped_image_gray)
-                        _, thresh_image = cv2.threshold(equalized_image, 65, 255, cv2.THRESH_BINARY_INV)
-                        cropped_image_3d = cv2.cvtColor(thresh_image, cv2.COLOR_GRAY2RGB)
-                        resized_image = cv2.resize(cropped_image_3d, (128, 32))
-                        
-                        pixel_values = self.processor_plate(resized_image, return_tensors="pt").pixel_values
-                        generated_ids = self.model_plate.generate(pixel_values)
-                        generated_text = self.processor_plate.batch_decode(generated_ids, skip_special_tokens=True)[0]
-                        
-                        generated_province, _ = self.get_closest_province(generated_text)
-                        data["raw_province"] = generated_text
-                        data["province"] = generated_province
+            # Process province detections
+            for result in province_results:
+                for box in result.boxes:
+                    confidence = float(box.conf)
+                    if confidence < self.CONF_THRESHOLD:
+                        continue
+
+                    x1, y1, x2, y2 = map(int, box.xyxy.flatten())
+                    cropped_image = image[y1:y2, x1:x2]
+
+                    # Draw rectangle on output image (blue for province)
+                    cv2.rectangle(output_image, (x1, y1), (x2, y2), (255, 0, 0), 2)
+
+                    # Process province using TrOCR
+                    cropped_image_gray = cv2.cvtColor(cropped_image, cv2.COLOR_BGR2GRAY)
+                    equalized_image = cv2.equalizeHist(cropped_image_gray)
+                    _, thresh_image = cv2.threshold(equalized_image, 65, 255, cv2.THRESH_BINARY_INV)
+                    cropped_image_3d = cv2.cvtColor(thresh_image, cv2.COLOR_GRAY2RGB)
+                    resized_image = cv2.resize(cropped_image_3d, (128, 32))
+                    
+                    pixel_values = self.processor_plate(resized_image, return_tensors="pt").pixel_values
+                    generated_ids = self.model_plate.generate(pixel_values)
+                    generated_text = self.processor_plate.batch_decode(generated_ids, skip_special_tokens=True)[0]
+                    
+                    generated_province, _ = self.get_closest_province(generated_text)
+                    data["raw_province"] = generated_text
+                    data["province"] = generated_province
 
             # Save the output image
             cv2.imwrite('output_detection.jpg', output_image)
@@ -170,13 +182,13 @@ def main():
     if uploaded_file is not None:
         # Create columns for side-by-side display
         col1, col2 = st.columns(2)
-
+        
         # Display original image
         with col1:
             st.subheader("Original Image")
             image = Image.open(uploaded_file)
             st.image(image, use_column_width=True)
-
+            
             # Convert PIL Image to OpenCV format for processing
             image_array = np.array(image)
             if len(image_array.shape) == 3 and image_array.shape[2] == 4:
@@ -196,7 +208,7 @@ def main():
 
                 # Clean up temporary input file
                 os.remove(temp_path)
-
+                
                 if results:
                     # Display results
                     st.subheader("Detection Results")
@@ -211,7 +223,7 @@ def main():
                             <p>Raw Province Text: {results['raw_province']}</p>
                         </div>
                         """, unsafe_allow_html=True)
-
+                    
                     # Display detection visualization
                     with col2:
                         st.subheader("Detection Visualization")
@@ -257,4 +269,4 @@ def main():
         """)
 
 if __name__ == "__main__":
-    main() 
+    main()

best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b1da8d9362a1005aa5b060b0ac53b4622677e753eded2893da10b6a69bc9fb7
+size 5468691

data.yaml

@@ -0,0 +1,52 @@
+train: /CarLicensePlate/iotproject-license-plate-3/train
+val: /CarLicensePlate/iotproject-license-plate-3/valid
+test: /CarLicensePlate/iotproject-license-plate-3/test
+nc: 47
+names: ['0', '1', '10', '11', '12', '13', '14', '15', '16', '17', '18', '19', '2', '20', '21', '22', '23', '24', '25', '26', '27', '28', '29', '3', '30', '31', '32', '33', '34', '35', '36', '37', '38', '39', '4', '40', '41', '42', '43', '44', '45', '46', '5', '6', '7', '8', '9']
+
+
+char_mapping:
+  '10': 'ก'
+  '11': 'ข'
+  '12': 'ค'
+  '13': 'ฆ'
+  '14': 'ง'
+  '15': 'จ'
+  '16': 'ฉ'
+  '17': 'ช'
+  '18': 'ฌ'
+  '19': 'ญ'
+  '20': 'ฎ'
+  '21': 'ฐ'
+  '22': 'ฒ'
+  '23': 'ณ'
+  '24': 'ด'
+  '25': 'ต'
+  '26': 'ถ'
+  '27': 'ท'
+  '28': 'ธ'
+  '29': 'น'
+  '30': 'บ'
+  '31': 'ผ'
+  '32': 'พ'
+  '33': 'ฟ'
+  '34': 'ภ'
+  '35': 'ม'
+  '36': 'ย'
+  '37': 'ร'
+  '38': 'ล'
+  '39': 'ว'
+  '40': 'ศ'
+  '41': 'ษ'
+  '42': 'ส'
+  '43': 'ห'
+  '44': 'ฬ'
+  '45': 'อ'
+  '46': 'ฮ'
+
+roboflow:
+  workspace: magarthai
+  project: iotproject-license-plate
+  version: 3
+  license: CC BY 4.0
+  url: https://universe.roboflow.com/magarthai/iotproject-license-plate/dataset/3

detect_plate.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:40d605afb93097eb60af1ea2bd33d5ae25ef778176f3a7e28be79add2d331890
+size 19188819