serialize depth then 3D so timeouts for 3D dont break whole system if user not want 3D

app.py

@@ -20,7 +20,7 @@ print(f"Timm version: {timm.__version__}")
 
 subprocess.run(["bash", "get_pretrained_models.sh"])
 
-@spaces.GPU(duration=20)
+@spaces.GPU(duration=30)
 def load_model_and_predict(image_path):
     device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
     model, transform = depth_pro.create_model_and_transforms()
@@ -195,7 +195,7 @@ def remove_thin_features(mesh, thickness_threshold=0.01):
     
     return mesh
 
-@spaces.GPU(duration=20)
+@spaces.GPU(duration=30)  # Increased duration to 30 seconds
 def regenerate_3d_model(depth_csv, image_path, focallength_px, simplification_factor, smoothing_iterations, thin_threshold):
     # Load depth from CSV
     depth = np.loadtxt(depth_csv, delimiter=',')
@@ -208,7 +208,7 @@ def regenerate_3d_model(depth_csv, image_path, focallength_px, simplification_fa
     print("regenerated!")
     return view_model_path, download_model_path
 
-@spaces.GPU(duration=20)
+@spaces.GPU(duration=30)
 def predict_depth(input_image):
     temp_file = None
     try:
@@ -238,18 +238,32 @@ def predict_depth(input_image):
         raw_depth_path = "raw_depth_map.csv"
         np.savetxt(raw_depth_path, depth, delimiter=',')
 
-        view_model_path, download_model_path = generate_3d_model(depth, temp_file, focallength_px)
-        print("Depth created - 3D model generated!")
-        return output_path, f"Focal length: {focallength_px:.2f} pixels", raw_depth_path, view_model_path, download_model_path, temp_file, focallength_px
+        print("Depth map created!")
+        return output_path, f"Focal length: {focallength_px:.2f} pixels", raw_depth_path, temp_file, focallength_px
     except Exception as e:
         import traceback
         error_message = f"An error occurred: {str(e)}\n\nTraceback:\n{traceback.format_exc()}"
         print(error_message)
-        return None, error_message, None, None, None, None, None
+        return None, error_message, None, None, None
     finally:
         if temp_file and os.path.exists(temp_file):
             os.remove(temp_file)
 
+@spaces.GPU(duration=30)
+def create_3d_model(depth_csv, image_path, focallength_px, simplification_factor, smoothing_iterations, thin_threshold):
+    try:
+        depth = np.loadtxt(depth_csv, delimiter=',')
+        view_model_path, download_model_path = generate_3d_model(
+            depth, image_path, focallength_px, 
+            simplification_factor, smoothing_iterations, thin_threshold
+        )
+        print("3D model generated!")
+        return view_model_path, download_model_path, "3D model created successfully!"
+    except Exception as e:
+        error_message = f"An error occurred during 3D model creation: {str(e)}"
+        print(error_message)
+        return None, None, error_message
+
 def get_last_commit_timestamp():
     try:
         timestamp = subprocess.check_output(['git', 'log', '-1', '--format=%cd', '--date=iso']).decode('utf-8').strip()
@@ -267,12 +281,10 @@ with gr.Blocks() as iface:
         "An enhanced demo that creates a textured 3D model from the input image and depth map.\n\n"
         "Forked from https://huggingface.co/spaces/akhaliq/depth-pro and model from https://huggingface.co/apple/DepthPro\n"
         "**Instructions:**\n"
-        "1. Upload an image.\n"
-        "2. The app will predict the depth map, display it, and provide the focal length.\n"
-        "3. Download the raw depth data as a CSV file.\n"
-        "4. View the generated 3D model textured with the original image.\n"
-        "5. Adjust parameters and click 'Regenerate 3D Model' to update the model.\n"
-        "6. Download the 3D model as an OBJ file if desired.\n\n"
+        "1. Upload an image to generate the depth map.\n"
+        "2. Click 'Generate 3D Model' to create the 3D visualization.\n"
+        "3. Adjust parameters and click 'Regenerate 3D Model' to update the model.\n"
+        "4. Download the raw depth data as a CSV file or the 3D model as an OBJ file if desired.\n\n"
         f"Last updated: {last_updated}"
     )
     
@@ -283,6 +295,8 @@ with gr.Blocks() as iface:
     focal_length = gr.Textbox(label="Focal Length")
     raw_depth_csv = gr.File(label="Download Raw Depth Map (CSV)")
     
+    generate_3d_button = gr.Button("Generate 3D Model")
+    
     with gr.Row():
         view_3d_model = gr.Model3D(label="View 3D Model")
         download_3d_model = gr.File(label="Download 3D Model (OBJ)")
@@ -293,22 +307,28 @@ with gr.Blocks() as iface:
         thin_threshold = gr.Slider(minimum=0.001, maximum=0.1, value=0.01, step=0.001, label="Thin Feature Threshold")
     
     regenerate_button = gr.Button("Regenerate 3D Model")
+    model_status = gr.Textbox(label="3D Model Status")
     
     # Hidden components to store intermediate results
-    hidden_depth_csv = gr.State()
     hidden_image_path = gr.State()
     hidden_focal_length = gr.State()
     
     input_image.change(
         predict_depth,
         inputs=[input_image],
-        outputs=[depth_map, focal_length, raw_depth_csv, view_3d_model, download_3d_model, hidden_image_path, hidden_focal_length]
+        outputs=[depth_map, focal_length, raw_depth_csv, hidden_image_path, hidden_focal_length]
+    )
+    
+    generate_3d_button.click(
+        create_3d_model,
+        inputs=[raw_depth_csv, hidden_image_path, hidden_focal_length, simplification_factor, smoothing_iterations, thin_threshold],
+        outputs=[view_3d_model, download_3d_model, model_status]
     )
     
     regenerate_button.click(
-        regenerate_3d_model,
+        create_3d_model,
         inputs=[raw_depth_csv, hidden_image_path, hidden_focal_length, simplification_factor, smoothing_iterations, thin_threshold],
-        outputs=[view_3d_model, download_3d_model]
+        outputs=[view_3d_model, download_3d_model, model_status]
     )
 
 # Launch the Gradio interface with sharing enabled