Update app.py

app.py

@@ -1,4 +1,4 @@
-import streamlit as st
+import gradio as gr
 from PIL import Image
 import torch
 from transformers import pipeline
@@ -7,140 +7,106 @@ import matplotlib.pyplot as plt
 from reportlab.pdfgen import canvas
 import io
 import folium
-from streamlit_folium import st_folium
-import cv2
 import tempfile
+import cv2
 import numpy as np
 
-st.set_page_config(page_title="InfraGuard Ultimate – Smart City AI", layout="wide")
-st.title("InfraGuard Ultimate – Live Smart City Incident Detection")
-
 # ---------------- Load Models ----------------
-@st.cache_resource
-def load_detection_model():
-    model = torch.hub.load('ultralytics/yolov5', 'yolov5s', pretrained=True)
-    return model
-
-@st.cache_resource
-def load_summarizer():
-    summarizer = pipeline("summarization", model="facebook/bart-large-cnn")
-    return summarizer
-
-detection_model = load_detection_model()
-summarizer = load_summarizer()
-
-# ---------------- Sidebar / Tabs ----------------
-tab1, tab2, tab3 = st.tabs(["Live Detection", "History", "About"])
+detection_model = torch.hub.load('ultralytics/yolov5', 'yolov5s', pretrained=True)
+summarizer = pipeline("summarization", model="facebook/bart-large-cnn")
 
 # ---------- Incident History ----------
-if "history" not in st.session_state:
-    st.session_state["history"] = pd.DataFrame(columns=["Image", "Objects", "Severity", "Summary", "Latitude", "Longitude"])
-
-# ---------------- Tab 1: Live Detection ----------------
-with tab1:
-    st.markdown("### Live Webcam / Upload Video or Image")
-    video_file = st.file_uploader("Upload Video", type=["mp4"])
-    uploaded_file = st.file_uploader("Upload Image", type=["jpg", "png", "jpeg"])
-    camera_input = st.camera_input("Or capture live image")
-
-    # Input GPS coordinates (simulate or from device)
-    latitude = st.number_input("Latitude", value=24.8607)  # default Karachi
-    longitude = st.number_input("Longitude", value=67.0011)
-
-    # Process Image
-    input_image = uploaded_file or camera_input
-    if input_image:
-        image = Image.open(input_image)
-        st.image(image, caption="Input Image", use_column_width=True)
-
-        results = detection_model(image)
-        st.image(results.render()[0], caption="Detected Objects", use_column_width=True)
-
-        detected_objects = results.pandas().xyxy[0]['name'].tolist()
-        st.markdown(f"**Detected Objects:** {', '.join(detected_objects)}")
-
-        severity = "Low"
-        if "fire" in detected_objects or "person" in detected_objects and len(detected_objects) > 15:
-            severity = "High"
-        elif "car" in detected_objects or "truck" in detected_objects:
-            severity = "Medium"
-        st.markdown(f"**Severity:** {severity}")
-
-        incident_text = f"Detected objects: {', '.join(detected_objects)}. Severity: {severity}."
-        summary = summarizer(incident_text, max_length=60, min_length=25, do_sample=False)
-        ai_summary = summary[0]['summary_text']
-        st.markdown("### AI-Generated Incident Summary")
-        st.write(ai_summary)
-
-        st.session_state["history"] = pd.concat([st.session_state["history"],
-                                                 pd.DataFrame([[input_image, ', '.join(detected_objects), severity, ai_summary, latitude, longitude]],
-                                                              columns=["Image", "Objects", "Severity", "Summary", "Latitude", "Longitude"])],
-                                                 ignore_index=True)
-
-        buffer = io.BytesIO()
-        c = canvas.Canvas(buffer)
-        c.drawString(100, 800, "InfraGuard Incident Report")
-        c.drawString(100, 780, f"Detected Objects: {', '.join(detected_objects)}")
-        c.drawString(100, 760, f"Severity: {severity}")
-        c.drawString(100, 740, f"AI Summary: {ai_summary}")
-        c.drawString(100, 720, f"Location: {latitude}, {longitude}")
-        c.save()
-        buffer.seek(0)
-        st.download_button("Download PDF Report", buffer, file_name="incident_report.pdf", mime="application/pdf")
-
-    # Process Video / Live Streaming
-    if video_file:
-        tfile = tempfile.NamedTemporaryFile(delete=False)
-        tfile.write(video_file.read())
-        cap = cv2.VideoCapture(tfile.name)
-
-        stframe = st.empty()
-        while cap.isOpened():
-            ret, frame = cap.read()
-            if not ret:
-                break
-            # Convert to RGB
-            frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
-            # Convert to PIL for YOLO
-            pil_img = Image.fromarray(frame_rgb)
-            results = detection_model(pil_img)
-            frame_result = np.squeeze(results.render())
-            stframe.image(frame_result, channels="RGB")
-        cap.release()
-
-# ---------------- Tab 2: History ----------------
-with tab2:
-    st.markdown("### Incident History")
-    st.dataframe(st.session_state["history"])
-
-    if not st.session_state["history"].empty:
-        all_objects = ','.join(st.session_state["history"]["Objects"].tolist()).split(',')
-        object_counts = pd.Series(all_objects).value_counts()
-        st.markdown("### Most Frequent Detected Objects")
-        fig, ax = plt.subplots()
-        object_counts.plot(kind='bar', ax=ax)
-        st.pyplot(fig)
-
-        st.markdown("### Incident Locations Map")
-        m = folium.Map(location=[24.8607, 67.0011], zoom_start=12)
-        for idx, row in st.session_state["history"].iterrows():
-            folium.Marker(
-                location=[row["Latitude"], row["Longitude"]],
-                popup=f"{row['Objects']} | Severity: {row['Severity']}",
-                icon=folium.Icon(color="red" if row["Severity"]=="High" else "orange" if row["Severity"]=="Medium" else "green")
-            ).add_to(m)
-        st_data = st_folium(m, width=700, height=500)
-
-# ---------------- Tab 3: About ----------------
-with tab3:
-    st.markdown("""
-    **InfraGuard Ultimate – Smart City AI System**  
-    - Real-time live webcam / video detection  
-    - Object detection using YOLOv5  
-    - AI-generated incident summary using BART  
-    - Severity scoring & real-time alerts  
-    - Incident history with analytics  
-    - Interactive map showing all incidents  
-    - Downloadable PDF incident reports  
-    - Fully deployable on Hugging Face Spaces
-    """)
+history_df = pd.DataFrame(columns=["Image", "Objects", "Severity", "Summary", "Latitude", "Longitude"])
+
+# ---------------- Helper Functions ----------------
+def process_image(img, latitude=24.8607, longitude=67.0011):
+    global history_df
+    image = Image.fromarray(img) if isinstance(img, np.ndarray) else img
+    results = detection_model(image)
+    detected_objects = results.pandas().xyxy[0]['name'].tolist()
+    
+    # Severity scoring
+    severity = "Low"
+    if "fire" in detected_objects or "person" in detected_objects and len(detected_objects) > 15:
+        severity = "High"
+    elif "car" in detected_objects or "truck" in detected_objects:
+        severity = "Medium"
+
+    # AI summary
+    incident_text = f"Detected objects: {', '.join(detected_objects)}. Severity: {severity}."
+    summary = summarizer(incident_text, max_length=60, min_length=25, do_sample=False)
+    ai_summary = summary[0]['summary_text']
+
+    # Save to history
+    history_df = pd.concat([history_df,
+                            pd.DataFrame([[image, ', '.join(detected_objects), severity, ai_summary, latitude, longitude]],
+                                         columns=["Image", "Objects", "Severity", "Summary", "Latitude", "Longitude"])],
+                           ignore_index=True)
+
+    # Prepare image with bounding boxes
+    output_image = np.squeeze(results.render())
+
+    # Prepare PDF
+    buffer = io.BytesIO()
+    c = canvas.Canvas(buffer)
+    c.drawString(100, 800, "InfraGuard Incident Report")
+    c.drawString(100, 780, f"Detected Objects: {', '.join(detected_objects)}")
+    c.drawString(100, 760, f"Severity: {severity}")
+    c.drawString(100, 740, f"AI Summary: {ai_summary}")
+    c.drawString(100, 720, f"Location: {latitude}, {longitude}")
+    c.save()
+    buffer.seek(0)
+
+    return output_image, ai_summary, severity, buffer
+
+def show_history():
+    global history_df
+    return history_df
+
+def show_map():
+    global history_df
+    m = folium.Map(location=[24.8607, 67.0011], zoom_start=12)
+    for idx, row in history_df.iterrows():
+        folium.Marker(
+            location=[row["Latitude"], row["Longitude"]],
+            popup=f"{row['Objects']} | Severity: {row['Severity']}",
+            icon=folium.Icon(color="red" if row["Severity"]=="High" else "orange" if row["Severity"]=="Medium" else "green")
+        ).add_to(m)
+    tmp_file = tempfile.NamedTemporaryFile(suffix=".html", delete=False)
+    m.save(tmp_file.name)
+    return tmp_file.name
+
+# ---------------- Gradio Interface ----------------
+with gr.Blocks() as demo:
+    gr.Markdown("## InfraGuard Ultimate – Smart City AI (Gradio Version)")
+
+    with gr.Tab("Live Detection"):
+        with gr.Row():
+            img_input = gr.Image(type="pil", label="Upload Image / Webcam")
+            latitude = gr.Number(value=24.8607, label="Latitude")
+            longitude = gr.Number(value=67.0011, label="Longitude")
+        with gr.Row():
+            detect_btn = gr.Button("Detect Incident")
+        with gr.Row():
+            output_image = gr.Image(label="Detection Output")
+        with gr.Row():
+            ai_summary_out = gr.Textbox(label="AI Incident Summary")
+            severity_out = gr.Textbox(label="Severity")
+        with gr.Row():
+            pdf_out = gr.File(label="Download PDF Report")
+
+        detect_btn.click(process_image,
+                         inputs=[img_input, latitude, longitude],
+                         outputs=[output_image, ai_summary_out, severity_out, pdf_out])
+
+    with gr.Tab("History & Analytics"):
+        hist_btn = gr.Button("Show Incident History")
+        hist_table = gr.Dataframe()
+        hist_btn.click(show_history, outputs=[hist_table])
+
+    with gr.Tab("Map"):
+        map_btn = gr.Button("Show Incident Map")
+        map_html = gr.HTML()
+        map_btn.click(show_map, outputs=[map_html])
+
+demo.launch()