Update app.py

app.py

@@ -7,28 +7,20 @@ import pickle
 import os
 import folium
 from folium import Rectangle
+import uuid
 
-# Disable GPU (optional)
-os.environ["CUDA_VISIBLE_DEVICES"] = "-1"
-
-# === Load model ===
+# Load model dan label encoder
 model = tf.keras.models.load_model("model_resnet50_jakarta_100epoch_240p_1088Grid_320train_80val_squared.h5")
-
-# === Load label encoder ===
 with open("label_encoder_resnet50_jakarta_100epoch_240p_1088Grid_320train_80val_squared.pkl", "rb") as f:
     label_encoder = pickle.load(f)
-
-# === Load class names ===
 class_names = label_encoder.classes_.tolist()
 
-# === Load BBOX Data ===
+# Load bounding box
 bbox_df = pd.read_csv("bbox_grid.csv")
-
 bbox_dict = {row['grid_id']: (row['min_lat'], row['min_lon'], row['max_lat'], row['max_lon']) for _, row in bbox_df.iterrows()}
 
 IMG_SIZE = (240, 240)
 
-# === Fungsi crop square ===
 def center_crop_square(img):
     h, w = img.shape[:2]
     min_dim = min(h, w)
@@ -36,7 +28,6 @@ def center_crop_square(img):
     start_y = h // 2 - min_dim // 2
     return img[start_y:start_y+min_dim, start_x:start_x+min_dim]
 
-# === Fungsi prediksi + generate peta ===
 def predict_and_map(image):
     image_cropped = center_crop_square(image)
     image_resized = cv2.resize(image_cropped, IMG_SIZE)
@@ -48,47 +39,42 @@ def predict_and_map(image):
     label = class_names[class_index]
     confidence = pred_probs[class_index]
 
-    # Ambil BBOX
-    if label in bbox_dict:
-        min_lat, min_lon, max_lat, max_lon = bbox_dict[label]
-    else:
-        return "Grid tidak ditemukan!", None
+    if label not in bbox_dict:
+        return "Grid tidak ditemukan!", None, None
 
-    # Buat peta
+    min_lat, min_lon, max_lat, max_lon = bbox_dict[label]
     center_lat = (min_lat + max_lat) / 2
     center_lon = (min_lon + max_lon) / 2
-    m = folium.Map(location=[center_lat, center_lon], zoom_start=15)
 
-    Rectangle(
-        bounds=[[min_lat, min_lon], [max_lat, max_lon]],
-        color="blue",
-        fill=True,
-        fill_opacity=0.4,
-    ).add_to(m)
+    # Buat peta
+    m = folium.Map(location=[center_lat, center_lon], zoom_start=15)
+    Rectangle([[min_lat, min_lon], [max_lat, max_lon]],
+              color="blue", fill=True, fill_opacity=0.4).add_to(m)
 
-    # Render HTML folium map
-    map_html = m.get_root().render()
+    # Simpan ke /tmp agar bisa diunduh di Spaces
+    filename = f"/tmp/map_{uuid.uuid4().hex}.html"
+    m.save(filename)
 
-    # Bungkus iframe
+    # Buat iframe preview
     iframe_html = f"""
-    <iframe srcdoc="{map_html.replace('"', '&quot;')}" 
+    <iframe srcdoc="{m.get_root().render().replace('"', '&quot;')}"
             style="width: 100%; height: 60vh; border: none;">
     </iframe>
     """
 
-    return f"Prediksi Grid: {label}\nConfidence: {confidence:.2%}", iframe_html
+    return f"Prediksi Grid: {label}\nConfidence: {confidence:.2%}", iframe_html, filename
 
-# === Gradio UI ===
-with gr.Blocks() as interface:
+# UI Gradio
+with gr.Blocks() as demo:
     with gr.Row():
         with gr.Column(scale=1):
             image_input = gr.Image(type="numpy", label="Upload Gambar Street View")
         with gr.Column(scale=2):
             text_output = gr.Textbox(label="Hasil Prediksi Grid", interactive=False)
-            map_output = gr.HTML(label="Peta Grid Prediksi")
+            map_output = gr.HTML(label="Peta Grid")
+            download_link = gr.File(label="Unduh Peta HTML")
 
-    # Event handling
-    image_input.change(fn=predict_and_map, inputs=image_input, outputs=[text_output, map_output])
+    image_input.change(fn=predict_and_map, inputs=image_input,
+                       outputs=[text_output, map_output, download_link])
 
-# Launch interface
-interface.launch()
+demo.launch()