Ajout modèle 4

app.py

@@ -2,12 +2,13 @@ import gradio as gr
 import tensorflow as tf
 import numpy as np
 from PIL import Image, ImageOps
+from tensorflow.keras.applications.efficientnet import preprocess_input
 
 # Charger les modèles (remplacez les chemins par vos fichiers de modèles)
 model1 = tf.keras.models.load_model('model_v2_Q1.keras')
 model2 = tf.keras.models.load_model('model_v2_Q2.keras')
 model3 = tf.keras.models.load_model('model_Q3.keras')
-#model4 = tf.keras.models.load_model('Model_4_Final.keras')
+model4 = tf.keras.models.load_model('model_Q4.keras')
 
 
 # Gère le formattage des images
@@ -22,6 +23,11 @@ def format_image_type2(image):
     image = np.array(image) / 255.0 # Normalisation
     return image_single_fix(image)
 
+def format_image_type3(image):
+    image = tf.image.resize(image, (128, 128)) # Redimensionner selon la taille attendue
+    image = preprocess_input(image) # Normalisation
+    return image_single_fix(image)
+
 def image_single_fix(image_array):
     return np.expand_dims(image_array, axis=0) # Ajouter batch dimension (demandé par tensorflow/keras qui demande à recevoir un groupe d'image plutôt que juste une seule)
 
@@ -42,6 +48,11 @@ models = {
         "model": model3,
         "type": "gender_age",
         "image_format": format_image_type2
+    },
+    "Modèle 4 (genre et âge) [Transfert d'apprentissage]": {
+        "model": model4,
+        "type": "age_gender",
+        "image_format": format_image_type3
     }
 }
 
@@ -64,10 +75,16 @@ def predict(image, model_name):
     elif model_type == "age_only":
         age_value = prediction[0][0]
         return_text = display_age_prediction(age_value)
-    else: # "gender_age"
+    elif model_type == "gender_age":
         gender_value = prediction[0][0][0]
         age_value = prediction[1][0][0]
         return_text = display_gender_prediction(gender_value) + "\n" + display_age_prediction(age_value)
+    elif model_type == "age_gender": # gender_age but prediction data order is reversed
+        gender_value = prediction[1][0][0]
+        age_value = prediction[0][0][0]
+        return_text = display_gender_prediction(gender_value) + "\n" + display_age_prediction(age_value)
+    else:
+        raise Exception(f"Unsupported model_type '{model_type}'")
     
     return f"{return_text}"