Update app.py

app.py

@@ -22,18 +22,23 @@ if not os.path.exists(extract_dir):
 
 model_tf = tf.saved_model.load(extract_dir)
 
+# --- Inspección de firma del modelo TensorFlow ---
+print("\n\n🔍 FIRMA DEL MODELO TENSORFLOW:")
+for key, func in model_tf.signatures.items():
+    print(f"Firma: {key}")
+    print("Entradas:", func.structured_input_signature)
+    print("Salidas:", func.structured_outputs)
+
 # Función helper para inferencia TensorFlow
 def predict_tf(img: Image.Image):
     try:
-        # Preprocesar imagen para TF: convertir a tensor float32, normalizar, añadir batch
-        img_resized = img.resize((224,224))  # ajusta según modelo
+        img_resized = img.resize((224,224))
         img_np = np.array(img_resized) / 255.0
-        if img_np.shape[-1] == 4:  # eliminar canal alfa si existe
+        if img_np.shape[-1] == 4:
             img_np = img_np[..., :3]
         img_tf = tf.convert_to_tensor(img_np, dtype=tf.float32)
-        img_tf = tf.expand_dims(img_tf, axis=0)  # batch dimension
+        img_tf = tf.expand_dims(img_tf, axis=0)
 
-        # Ejecutar modelo (suponiendo firma default)
         infer = model_tf.signatures["serving_default"]
         output = infer(img_tf)
         pred = list(output.values())[0].numpy()[0]
@@ -43,17 +48,14 @@ def predict_tf(img: Image.Image):
         print(f"Error en predict_tf: {e}")
         return np.zeros(len(CLASSES))
 
-# 🔹 Cargar modelo ViT desde Hugging Face
 MODEL_NAME = "ahishamm/vit-base-HAM-10000-sharpened-patch-32"
 feature_extractor = ViTImageProcessor.from_pretrained(MODEL_NAME)
 model_vit = ViTForImageClassification.from_pretrained(MODEL_NAME)
 model_vit.eval()
 
-# 🔹 Cargar modelos Fast.ai desde archivos locales
 model_malignancy = load_learner("ada_learn_malben.pkl")
 model_norm2000 = load_learner("ada_learn_skin_norm2000.pkl")
 
-# 🔹 Clases y niveles de riesgo
 CLASSES = [
     "Queratosis actínica / Bowen", "Carcinoma células basales",
     "Lesión queratósica benigna", "Dermatofibroma",
@@ -71,10 +73,7 @@ RISK_LEVELS = {
 
 def analizar_lesion_combined(img):
     try:
-        # Convertir imagen para Fastai
         img_fastai = PILImage.create(img)
-
-        # ViT prediction
         inputs = feature_extractor(img, return_tensors="pt")
         with torch.no_grad():
             outputs = model_vit(**inputs)
@@ -89,9 +88,8 @@ def analizar_lesion_combined(img):
         probs_vit = np.zeros(len(CLASSES))
 
     try:
-        # Fast.ai models
         pred_fast_malignant, _, probs_fast_mal = model_malignancy.predict(img_fastai)
-        prob_malignant = float(probs_fast_mal[1])  # 1 = maligno
+        prob_malignant = float(probs_fast_mal[1])
     except Exception as e:
         print(f"Error en Fast.ai malignancy: {e}")
         prob_malignant = 0.0
@@ -103,26 +101,24 @@ def analizar_lesion_combined(img):
         pred_fast_type = "Error"
 
     try:
-        # TensorFlow model prediction
         probs_tf = predict_tf(img)
         pred_idx_tf = int(np.argmax(probs_tf))
         confidence_tf = probs_tf[pred_idx_tf]
         if pred_idx_tf < len(CLASSES):
             pred_class_tf = CLASSES[pred_idx_tf]
         else:
-            pred_class_tf = f"Clase desconocida (índice {pred_idx_tf})"
+            pred_class_tf = f"Clase desconocida (\u00edndice {pred_idx_tf})"
     except Exception as e:
         print(f"Error en TensorFlow prediction: {e}")
         pred_class_tf = "Error"
         confidence_tf = 0.0
 
-    # Gráfico ViT
     colors_bars = [RISK_LEVELS[i]['color'] for i in range(7)]
     fig, ax = plt.subplots(figsize=(8, 3))
     ax.bar(CLASSES, probs_vit*100, color=colors_bars)
     ax.set_title("Probabilidad ViT por tipo de lesión")
     ax.set_ylabel("Probabilidad (%)")
-    ax.set_xticks(np.arange(len(CLASSES)))  # evita warning
+    ax.set_xticks(np.arange(len(CLASSES)))
     ax.set_xticklabels(CLASSES, rotation=45, ha='right')
     ax.grid(axis='y', alpha=0.2)
     plt.tight_layout()
@@ -133,7 +129,6 @@ def analizar_lesion_combined(img):
     img_b64 = base64.b64encode(img_bytes).decode("utf-8")
     html_chart = f'<img src="data:image/png;base64,{img_b64}" style="max-width:100%"/>'
 
-    # Informe HTML
     informe = f"""
     <div style="font-family:sans-serif; max-width:800px; margin:auto">
     <h2>🧪 Diagnóstico por 4 modelos de IA</h2>
@@ -145,10 +140,9 @@ def analizar_lesion_combined(img):
         <tr><td>🔬 TensorFlow (saved_model)</td><td><b>{pred_class_tf}</b></td><td>{confidence_tf:.1%}</td></tr>
     </table>
     <br>
-    <b>🩺 Recomendación automática:</b><br>
+    <b>🧪 Recomendación automática:</b><br>
     """
 
-    # Recomendación basada en ViT + malignidad (podrías adaptar aquí según TF)
     cancer_risk_score = sum(probs_vit[i] * RISK_LEVELS[i]['weight'] for i in range(7))
     if prob_malignant > 0.7 or cancer_risk_score > 0.6:
         informe += "🚨 <b>CRÍTICO</b> – Derivación urgente a oncología dermatológica"