Update app.py

app.py

@@ -6,34 +6,123 @@ import numpy as np
 
 # function which is returning the number of object detected 
 def number_object_detected(image):
-    custom_model = YOLO("best.pt")
-    results = custom_model(image)
+    
+    custom_model = YOLO('D:/COMSATS/FYP/Coding Part/ModelMaking/Computer_Vision_Part/3rd Try/runs/detect/train9/weights/best.pt') # custome yolo model path
+    results = custom_model(image,verbose= False)
 
     dic = results[0].names
-    cls = results[0].boxes.cls.cpu().numpy()
+    classes = results[0].boxes.cls.cpu().numpy()
+    probability = results[0].boxes.conf
+
     class_count = {}
-    unique_elements, counts = np.unique(cls, return_counts=True)
+    unique_elements, counts = np.unique(classes, return_counts=True)
     for e , count in zip(unique_elements,counts):
         a = dic[e]
         class_count[a] = count
-        print(f"the car has {count} {a}")
+    return class_count
+
+
+def car_detection_and_Cropping(image_path):
+    simple_yolo = YOLO('yolov8m.pt')
+    r = simple_yolo(image_path,verbose = False)
     
-    return class_count , results
 
+    names = r[0].names
+    boxes = r[0].boxes.xyxy.cpu().numpy().astype(int)
+    classes = set(r[0].boxes.cls.cpu().numpy())
+    classes2 =  [names[i] for i in classes]
+
+    # checking if the detected object is the car or not 
+    # if it is car then crop if not then pass the image as it is
+    if boxes.size != 0 and 'car' in classes2:
+      
+      area = []
+      for x1, y1, x2, y2 in boxes:
+          area.append((x2 - x1) * (y2 - y1))
+      max_index, max_a = max(enumerate(area), key=lambda x: x[1])
+      
+
+      # Load the image using OpenCV
+      image = cv2.imread(image_path)
+      
+      
+      # Crop the image
+      crop_image = image[boxes[max_index][1]:boxes[max_index][3], boxes[max_index][0]:boxes[max_index][2]]
+
+      # passing the crop image to the detection model
 
+      class_c = number_object_detected(crop_image)
+    else:
+      class_c = number_object_detected(image_path)
+    return class_c
+
+## this function will take the image url and call all the related functions
+def estimate_condition(detections):
+    print("Detedtion list",detections)
+    max_possible_score = sum(severity_points.values())  # Assuming all types of damage detected
+    score = calculate_condition_score(detections)
+    normalized_score = normalize_score(score, max_possible_score)
+    
+    if normalized_score <= 2:  # If score is low, condition is Excellent
+        print("Condition Excellent")
+        return "Excellent"
+    elif (normalized_score >2 and normalized_score <=7):  # If score is moderately low, condition is Good
+        print("Condition Good")
+        return "Good"
+    elif (normalized_score >7 and normalized_score <15):  # If score is moderate, condition is Fair
+        print("Condition Fair")
+        return "Fair"
+    elif (normalized_score >15 and normalized_score<=20):  # If score is moderately high, condition is Poor
+        print("Condition Poor")
+        return "Poor"
+    else:  # If score is high, condition is Very Poor
+        print("Condition Very Poor")
+        return "Very Poor"
+
+## loading the model
+with open('Price_prediction_decision_tree.pkl', 'rb') as file:
+    loaded_pipe_lr = pickle.load(file)
+    
 # Define function to process input image and return annotated image
-def detect_objects(input_image):
-    cls , result = number_object_detected(input_image)
-    # Plot the results and return annotated image
-    for r in result:
-        im_array = r.plot(pil = True)  # plot a BGR numpy array of predictions
-        array = im_array[..., ::-1]  # Convert BGR to RGB PIL image
-        plt.axis("off")
-        plt.imshow(array)
-        plt.show()
+def process_data(files,car_brand, car_name, model_year, mileage, city_registered, color, engine_c, trans, fuel_type, Cate):
+    print(car_brand, car_name, model_year, mileage, city_registered, color, engine_c, trans, fuel_type,  Cate)
+    # Process uploaded filespr
+    print(files)
+    file_names = [f for f in files[0]]
+    print(file_names)
+    # Process text input
+    damage_dic = {}
     
-    return array , cls
+    for i in file_names:
+          data = car_detection_and_Cropping(i)
+          print(data)
+          for key in data.keys():
+                      if key in damage_dic:
+                        damage_dic[key] += data[key]
+                      else:
+                          damage_dic[key] = data[key]
+    condition = estimate_condition(damage_dic)
+    #[[2022,46000 , 'punjba','white',1500,'changan','alsvin','automatic','petrol','excellent','sedan']])
+    price = loaded_pipe_lr.predict([[model_year,mileage,city_registered,color,engine_c,car_brand,car_name,trans,fuel_type,condition,Cate]])
+    if price[0] >= 100:
+          price[0] = price[0]/100
+    
+    return (damage_dic,condition , str(price[0])+'lacs')
+    
+
 
-# Define Gradio interface
-interface = gr.Interface(fn=detect_objects, inputs=gr.Image(type= 'pil', label='Upload Image of Car'), outputs=[gr.Image(),gr.Textbox(label="Number of Objects detected ")], title=" 🚘Car Scratch and Dent Detection")
-interface.launch()
+years_list = list(range(2024, 1899, -1))
+gr.Interface(fn = process_data,
+             inputs=[gr.Gallery(label="Upload Files", type="pil"),
+                  gr.Dropdown(['suzuki','toyota','honda','kia'], label='Brand'),
+                      gr.Textbox(lines=1, label="Car Name"),
+                      gr.Dropdown(choices=years_list, label='Model Year'),
+                      gr.Number(label="Mileage Km"),
+                      gr.Textbox(lines=1, label="City Register"),
+                      gr.Textbox(lines=1, label="Color"),
+                      gr.Number(label="Engine Capacity in CC"),
+                      gr.Radio(["automatic", "manual"], label="Transmission Type"),
+                     # gr.Radio(["imported", "local"], label="Assembly Type"),
+                      gr.Radio(["hybrid", "petrol",'diesel'], label="Fuel Type"),
+                      gr.Radio(["hatchback", "sedan",'suv','croosover'], label="Category")],
+                    outputs=[gr.Textbox(label="Damage List"),gr.Textbox(label="Condition"),gr.Textbox(label="Predicted Price")]).launch()