Update app.py

app.py

@@ -1,34 +1,40 @@
 import gradio as gr
-from transformers import DPTImageProcessor, DPTForDepthEstimation
 import torch
 import numpy as np
 from PIL import Image
 from pathlib import Path
 from depth_viewer import depthviewer2html
 
-feature_extractor = DPTImageProcessor.from_pretrained("Intel/dpt-large")
-model = DPTForDepthEstimation.from_pretrained("Intel/dpt-large")
+# Don't load models at module level
+_model = None
+_feature_extractor = None
+
+def get_model():
+    global _model, _feature_extractor
+    if _model is None:
+        from transformers import DPTImageProcessor, DPTForDepthEstimation
+        _feature_extractor = DPTImageProcessor.from_pretrained("Intel/dpt-large")
+        _model = DPTForDepthEstimation.from_pretrained("Intel/dpt-large")
+    return _feature_extractor, _model
 
 def process_image(image_path):
     if image_path is None:
         return ""
     
+    feature_extractor, model = get_model()
+    
     image_path = Path(image_path)
     image = Image.open(image_path)
     
-    # if wider than 512 pixels let's resample to keep it performant
     if image.size[0] > 512:
         image = image.resize((512, int(512 * image.size[1] / image.size[0])), Image.Resampling.LANCZOS)
     
-    # prepare image for the model
     inputs = feature_extractor(images=image, return_tensors="pt")
     
-    # forward pass
     with torch.no_grad():
         outputs = model(**inputs)
         predicted_depth = outputs.predicted_depth
     
-    # interpolate to original size
     prediction = torch.nn.functional.interpolate(
         predicted_depth.unsqueeze(1),
         size=image.size[::-1],
@@ -42,22 +48,21 @@ def process_image(image_path):
     return depthviewer2html(image, depth)
 
 title = "3D Visualization of Depth Maps Generated using MiDaS"
-description = "Improved 3D interactive depth viewer using Three.js embedded in a Gradio app."
+description = "Improved 3D interactive depth viewer using Three.js"
 
 with gr.Blocks(css="#depth-viewer { height: 600px; }") as demo:
     gr.Markdown(f"# {title}")
     gr.Markdown(description)
     
-    with gr.Row():
-        input_image = gr.Image(type="filepath", label="Input Image")
-    
+    input_image = gr.Image(type="filepath", label="Input Image")
     output_html = gr.HTML(label="Depth Viewer", elem_id="depth-viewer")
     
     input_image.change(fn=process_image, inputs=input_image, outputs=output_html)
     
     gr.Examples(
         examples=[["examples/owl1.jpg"], ["examples/marsattacks.jpg"], ["examples/kitten.jpg"]],
-        inputs=input_image
+        inputs=input_image,
+        cache_examples=False
     )
 
 demo.launch(server_name="0.0.0.0")