Update screens/predict.py

screens/predict.py

@@ -1,125 +1,133 @@
 import streamlit as st
-import pandas as pd
-import numpy as np
+import json
 from autogluon.multimodal import MultiModalPredictor
 from autogluon.tabular import TabularPredictor
+import pandas as pd
+from geopy.geocoders import GoogleV3
+import os
+import tempfile
+
+def predict_page():
+    # Add a title with an icon
+    st.image("icon.png", width=80)
+    st.title("💵 Property Price Estimator")
+
+    # User role selection (buyer or seller)
+    user_role = st.selectbox("Select Your Role", options=["Buyer", "Seller"])
+
+    # Load geocoder and model based on user role
+    @st.cache(allow_output_mutation=True)
+    def load_geocoder():
+        return GoogleV3(api_key=os.environ.get("GOOGLE_MAP_API_KEY"))
 
-# Define icons
-seller_icon = "🏡"
-buyer_icon = "🔍"
-submit_icon = "📝"
-predict_icon = "🔮"
-
-# Initialize df as a global variable
-df = None
-
-def predict_price():
-    global df  # Declare df as a global variable
-    # Set the title and subheader
-    st.title("Real Estate Price Prediction")
-    st.subheader("Choose your role and provide property details")
-
-    # User role selection
-    option = st.selectbox("Who are you?", ['Seller', 'Buyer'], index=0)
-
-    if option == "Seller":
-        st.subheader(f"{seller_icon} Seller Information")
-        with st.spinner("Loading model..."):
-            predictor = MultiModalPredictor.load("C:/Users/duong/OneDrive/Desktop/mm-nlp-image-transformer")
-        st.success("Done")
-        description = st.text_area("Property Description", help="Describe your property")
-        title = st.text_input("Property Title", help="Enter a title for your property")
+    geocoder = load_geocoder()
+    if user_role == "Buyer":
+        model_path = "models/mm-nlp-image-transformer/"
     else:
-        st.subheader(f"{buyer_icon} Buyer Information")
-        with st.spinner("Loading model..."):
-            predictor = TabularPredictor.load("C:/Users/duong/OneDrive/Desktop/tabular", require_py_version_match=False)
-        st.success("Done")
-
-    # Property details input
-    area = st.number_input("Property Area (square meters)", min_value=1)
-    location = st.text_input("Property Location", help="Enter the location of the property")
-    city_code = st.text_input("City Code", help="Enter the city code")
-    district = st.text_input("District", help="Enter the district name")
-    bedroom = st.slider("Number of Bedrooms", min_value=1, max_value=10, value=5, step=1)
-    bathroom = st.slider("Number of Bathrooms", min_value=1, max_value=10, value=2, step=1)
-
-    # Submit button to create the DataFrame
-    submitted = st.button(f"{submit_icon} Submit")
-
-    # Create a DataFrame from user inputs
-    if submitted: 
-        if area and location and city_code and district and bedroom and bathroom:
-            if option == "Seller":  
-                if (not description or not title):
-                    st.error("Please fill in both Description and Title fields for Sellers.")
-                else:
-                    data = {
-                        "Price": np.nan,
-                        "Area": [area],
-                        "Location": [location],
-                        "Time stamp": np.nan,
-                        "Certification status": np.nan,
-                        "Direction": np.nan,
-                        "Bedrooms": [bedroom],
-                        "Bathrooms": [bathroom],
-                        "Front width": np.nan,
-                        "Floor": np.nan,
-                        "Image URL": np.nan,
-                        "Road width": np.nan,
-                        "City_code": [city_code],
-                        "DistrictId": [district],
-                        "Balcony_Direction": np.nan,
-                        "Longitude": np.nan,
-                        "Lattitude": np.nan,
-                        "Description": [description],
-                        "Title": [title]
-                    }
-                    df = pd.DataFrame(data)
-                    st.write(f"{seller_icon} Input Data:")
-                    st.dataframe(df)
-            elif option == "Buyer":
-                data = {
-                        "Price": np.nan,
-                        "Area": [area],
-                        "Location": [location],
-                        "Time stamp": np.nan,
-                        "Certification status": np.nan,
-                        "Direction": np.nan,
-                        "Bedrooms": [bedroom],
-                        "Bathrooms": [bathroom],
-                        "Front width": np.nan,
-                        "Floor": np.nan,
-                        "Image URL": np.nan,
-                        "Road width": np.nan,
-                        "City_code": [city_code],
-                        "DistrictId": [district],
-                        "Balcony_Direction": np.nan,
-                        "Longitude": np.nan,
-                        "Lattitude": np.nan
-                    }
-                df = pd.DataFrame(data)
-                st.write(f"{buyer_icon} Input Data:")
-                st.dataframe(df)
-        else:
-            st.error("Please fill in all fields to have a better prediction!")
-
-    # Prediction button (enabled only when data has been submitted)
-    if st.button(f"{predict_icon} Predict"):
-        with st.spinner("Loading..."):
-            # Perform predictions and calculations here
-            predictions = predictor.predict(df.drop(columns="Price"))
-            st.success(f"Predicted Price: {predictions[0]:,.0f} VND")
-
-            scores = predictor.evaluate(
-                df,
-                metrics=[
-                    "mean_squared_error",
-                    "r2",
-                ],
-            )
-
-            st.subheader("Model Evaluation Metrics:")
-            for metric, score in scores.items():
-                st.write(f"{metric}: {score:.2f}")
+        model_path = "models/tabular/"
+
+    @st.cache(allow_output_mutation=True)
+    def load_mm_text_model():
+        if user_role == "Seller":
+            return MultiModalPredictor.load(model_path, verbosity=0)
+        elif user_role == "Buyer":
+            return TabularPredictor.load(model_path, verbosity=0)
+
+    mm_text_predictor = load_mm_text_model()
+
+    # Common user input fields
+    st.header("Location Details")
+    city_map = json.load(open("city-map.json"))
+    city = st.selectbox("Choose city 🏙️", options=list(city_map.values()))
+    city_district_map = json.load(open("city-district-map.json"))
+    district = st.selectbox("Choose district 🏘️", options=list(city_district_map[city].values()))
+    location = st.text_input("Enter precise location 📍")
+
+    st.header("Property Specifications")
+    area = st.number_input("Area (m2) 📏", min_value=1.0)
+    bedrooms = st.number_input("Number of bedrooms 🛏️", min_value=1, value=1)
+    bathrooms = st.number_input("Number of bathrooms 🚽", min_value=1, value=1)
+    floors = st.number_input("Number of floors 🏢", min_value=1, value=1)
+    front_width = st.number_input("Front width (m) 📏", min_value=0.0, value=0.0, step=0.1)
+    road_width = st.number_input("Road width (m) 🚗", min_value=0.0, value=0.0, step=0.1)
+
+    st.header("Additional Details")
+    timestamp = st.date_input("Date posted 📅")
+    cert_status = st.selectbox("Certification status 📜", options=["Không có", "hợp đồng", "sổ đỏ / sổ hồng"])
+    direction = st.selectbox("Direction 🧭", options=["Không có", "Tây - Nam", "Đông - Nam", "Đông - Bắc", "Tây - Bắc", "Nam", "Tây", "Bắc", "Đông"])
+    balcony_direction = st.selectbox("Balcony direction 🌞", options=["Không có", "Tây - Nam", "Đông - Nam", "Đông - Bắc", "Tây - Bắc", "Nam", "Tây", "Bắc", "Đông"])
+
+    # Description and image upload (conditional for sellers)
+    if user_role == "Seller":
+        description = st.text_area("Description ✍️")
+        title = description.split(".", maxsplit=1)[0]
+
+        uploaded_image = st.file_uploader("Upload an image 📷")
+        image_tmp = None
+        if uploaded_image:
+            image_tmp = tempfile.NamedTemporaryFile(suffix=uploaded_image.name)
+            image_tmp.write(uploaded_image.read())
+            st.image(uploaded_image, caption='Uploaded Image', use_column_width=True)
+    else:
+        description = ""
+        title = ""
+
+    # Calculate latitude and longitude
+    location = (location + ", " if location else "") + city + ", " + district
+    geocode_result = geocoder.geocode(query=location, region="vn", language="vi")
+    latitude = geocode_result.latitude
+    longitude = geocode_result.longitude
+
+    # Display location map
+    st.map([(latitude, longitude)], zoom=15)
+
+    # Create a button with an icon to get estimated price
+    if st.button("Get Estimated Price with Text 💰"):
+        if user_role == "Seller":
+            input_data = {
+                "Title": title,
+                "Area": area,
+                "Location": location,
+                "Time stamp": timestamp,
+                "Certification status": cert_status,
+                "Direction": direction,
+                "Bedrooms": bedrooms,
+                "Bathrooms": bathrooms,
+                "Front width": front_width or float("nan"),
+                "Floor": floors,
+                "Description": description,
+                "Image URL": image_tmp.name if image_tmp else None,
+                "Road width": road_width or float("nan"),
+                "City_code": city,
+                "DistrictId": district,
+                "Lattitude": latitude,
+                "Longitude": longitude,
+                "Balcony_Direction": balcony_direction,
+            }
+        elif user_role == "Buyer":
+            input_data = {
+                "Area": area,
+                "Location": location,
+                "Time stamp": timestamp,
+                "Certification status": cert_status,
+                "Direction": direction,
+                "Bedrooms": bedrooms,
+                "Bathrooms": bathrooms,
+                "Front width": front_width or float("nan"),
+                "Floor": floors,
+                "Image URL": image_tmp.name if image_tmp else None,
+                "Road width": road_width or float("nan"),
+                "City_code": city,
+                "DistrictId": district,
+                "Lattitude": latitude,
+                "Longitude": longitude,
+                "Balcony_Direction": balcony_direction,
+            }
+
+        input_df = pd.DataFrame([input_data])
+        predicted_price = mm_text_predictor.predict(input_df, as_pandas=False).item()
 
+        # Display the estimated price with an icon
+        st.subheader("Estimated Price 💲")
+        st.write(f"{predicted_price * 1e6:,.0f} VND")