Update app.py

app.py

@@ -11,16 +11,18 @@ import numpy as np
 
 model = tf.keras.models.load_model('mnist.keras') # a modell betoltese
 
-def preprocess_image(img):     # Preprocess the input image
-    if len(img.shape) == 3 and img.shape[-1] == 3:
-        img = tf.image.rgb_to_grayscale(img).numpy().squeeze()
+def preprocess_image(img):     # Potenciális RGB 2 gray konverzió
+    if len(img.shape) == 3 and img.shape[-1] == 3: # ha szines
+        img = tf.image.rgb_to_grayscale(img).numpy().squeeze() 
+        # hatekony gray conv. > np array lesz belőle
 
-    img = tf.image.resize(img[None, ..., None], (28, 28)).numpy().squeeze()
+    img = tf.image.resize(img[None, ..., None], (28, 28)).numpy().squeeze() # resize
+    # dimenziót is váltogatunk, a tf.image.resizenak kell ez
 
-    if img.mean() > 127:
+    if img.mean() > 127: # szin invert
         img = 255 - img
 
-    img = img.astype('float32') / 255.0  # Normalize
+    img = img.astype('float32') / 255.0  # Normalizál
     img = img.reshape(784)  # Flatten the image
 
     return img
@@ -29,7 +31,7 @@ def predict_digit(img):      # Predict the digit in the image
     processed_img = preprocess_image(img)
     processed_img = np.expand_dims(processed_img, axis=0)  # Add batch dimension
     prediction = model.predict(processed_img)[0]
-    return {str(i): float(prediction[i]) for i in range(10)}
+    return {str(i): float(prediction[i]) for i in range(10)} # Gradio comaptible kiiras
 
 ## Fonti a mukodteto funkcio kodok
 ## Alabb a Gradio interface resz