Update app.py

app.py

@@ -1,8 +1,9 @@
 import gradio as gr
 import torch
 from transformers import AutoFeatureExtractor, AutoModelForImageClassification
-from PIL import Image, ImageDraw, ImageFont
+from PIL import Image
 import numpy as np
+import cv2
 
 class ThyroidTumorClassifierApp:
     def __init__(self):
@@ -32,46 +33,44 @@ class ThyroidTumorClassifierApp:
         # Predicted class label
         predicted_label = class_labels[predicted_class]
         
-        # Add information to the output image
+        # Add information to the output image using OpenCV
         output_image_with_info = self.add_info_to_image(image, predicted_label, probabilities)
         
-        # Convert the output image to PIL format
-        output_pil_image = Image.fromarray(output_image_with_info)
-        
-        # Return the modified output image as PIL image
-        return output_pil_image
+        # Return the modified output image as an array
+        return output_image_with_info
 
     def add_info_to_image(self, image, predicted_label, probabilities):
-        # Convert the image to Pillow image format
-        image_pil = Image.fromarray(image)
-        
-        # Create a drawing object to add text to the image
-        draw = ImageDraw.Draw(image_pil)
+        # Convert the image to RGB format
+        image_rgb = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
         
-        # Choose the font and text size
-        font = ImageFont.truetype("arial.ttf", 20)
+        # Add the predicted class label and probabilities to the image using OpenCV
+        font = cv2.FONT_HERSHEY_SIMPLEX
+        font_scale = 0.6
+        font_thickness = 1
+        text_color = (255, 255, 255)
+        text_position = (10, 30)
         
-        # Add the predicted class label to the image
-        draw.text((10, 10), f"Classe Prevista: {predicted_label}", fill="white", font=font)
+        # Add predicted class label
+        cv2.putText(image_rgb, f"Classe Prevista: {predicted_label}", text_position, font, font_scale, text_color, font_thickness)
         
-        # Add the probabilities to the image
-        for i, prob in enumerate(probabilities):
-            y_offset = 40 + i * 30
+        # Add class probabilities
+        for i, prob in enumerate(probabilities[0]):
+            y_offset = 60 + i * 30
             class_name = f"Classe {i}:"
             probability = f"{prob:.2f}"
-            draw.text((10, y_offset), f"{class_name} {probability}", fill="white", font=font)
+            cv2.putText(image_rgb, f"{class_name} {probability}", (10, text_position[1] + y_offset), font, font_scale, text_color, font_thickness)
         
-        # Convert back to numpy format
-        image_with_info = np.array(image_pil)
+        # Convert back to BGR format for display
+        output_image = cv2.cvtColor(image_rgb, cv2.COLOR_RGB2BGR)
         
-        return image_with_info
+        return output_image
 
     def run_interface(self):
         # Create a Gradio interface
         input_interface = gr.Interface(
             fn=self.classify_image,
             inputs=gr.inputs.Image(),
-            outputs=gr.outputs.Image(type="pil"),  # Use type="pil" for Pillow image format
+            outputs=gr.outputs.Image(),
             title="Tumor da Tireoide Classificação",
             description="Faça o upload de uma imagem de um tumor da tireoide para classificação. A saída inclui o rótulo da classe prevista e as probabilidades com informações adicionais.",
         )