Create app.py

app.py

@@ -0,0 +1,40 @@
+import os
+import onnxruntime
+import gradio as gr
+import numpy as np
+from PIL import Image
+
+onnx_model_path = "sarcoloring.onnx"
+sess = onnxruntime.InferenceSession(onnx_model_path)
+
+def predict(input_image):
+   
+    input_image = input_image.resize((256, 256))  
+    input_image = np.array(input_image).transpose(2, 0, 1) 
+    input_image = input_image.astype(np.float32) / 255.0 
+    input_image = (input_image - 0.5) / 0.5              
+    input_image = np.expand_dims(input_image, axis=0)  
+    
+    # Run the model
+    inputs = {sess.get_inputs()[0].name: input_image}
+    output = sess.run(None, inputs)
+    
+    
+    output_image = output[0].squeeze().transpose(1, 2, 0)  
+    output_image = (output_image + 1) / 2 # [0,1]
+    output_image = (output_image * 255).astype(np.uint8)  
+    
+    return Image.fromarray(output_image)
+
+
+example_images = [[os.path.join("examples", fname)] for fname in os.listdir("examples")]
+
+
+iface = gr.Interface(fn=predict, 
+                     inputs=gr.Image(type="pil"), 
+                     outputs=gr.Image(type="pil"),
+                     examples=example_images
+                     )
+
+
+iface.launch()