Update app.py

app.py

@@ -11,6 +11,7 @@ import os
 import json
 from openvino.runtime import Core  # Assuming you're using OpenVINO
 from tf_post_processing import non_max_suppression  # Assuming this is defined elsewhere
+from PIL import Image
 
 # Load the OpenVINO model
 classification_model_xml = "./model/best.xml"
@@ -39,6 +40,10 @@ label_to_class_text = {
 
 # Function to perform inference
 def predict_image(image):
+    # Convert the Pillow image to a NumPy array and BGR format for OpenCV
+    image = np.array(image.convert("RGB"))
+    image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
+    
     # Resize, preprocess, and reshape the input image
     img_size = 960
     resized_image = cv2.resize(image, (img_size, img_size)) / 255.0
@@ -89,20 +94,14 @@ def predict_image(image):
                 }
             })
 
-    # Save the processed image and JSON file
+    # Save the processed image
     output_image_path = os.path.join(output_image_folder, "result_image.jpg")
     cv2.imwrite(output_image_path, image)
 
-    output_json_path = os.path.join(output_json_folder, "predictions.json")
-    with open(output_json_path, 'w') as f:
-        json.dump(predictions, f, indent=4)
-
-    return image, predictions
+    # Convert predictions to a formatted string
+    predictions_str = json.dumps(predictions, indent=4)
 
-# Set up Gradio interface with sample images and image upload
-def gradio_interface(image):
-    output_image, predictions = predict_image(image)
-    return output_image, json.dumps(predictions, indent=4)
+    return image, predictions_str
 
 # Define sample images for user convenience
 sample_images = [
@@ -111,11 +110,16 @@ sample_images = [
     "./sample/10_12.jpg"
 ]
 
+# Set up Gradio interface
+def gradio_interface(image):
+    output_image, predictions_str = predict_image(image)
+    return output_image, predictions_str
+
 # Create the Gradio Interface
 gr_interface = gr.Interface(
     fn=gradio_interface,
-    inputs=gr.inputs.Image(label="Upload or Select an Image", type="numpy", sample_images=sample_images),
-    outputs=[gr.outputs.Image(label="Result Image"), gr.outputs.Textbox(label="Predictions JSON")],
+    inputs=gr.Image(label="Upload or Select an Image", type="pil", sample_images=sample_images),
+    outputs=[gr.Image(label="Result Image"), gr.Textbox(label="Predictions JSON")],
     title="OpenVINO Model Inference with Gradio",
     description="Upload an image or select a sample image to get model predictions with bounding boxes and probabilities."
 )