Update app.py

app.py

@@ -8,6 +8,8 @@ import numpy as np
 import gradio as gr
 import io
 import base64
+from torchvision import transforms
+from efficientnet_pytorch import EfficientNet
 
 # --- Cargar modelo ViT preentrenado fine‑tuned HAM10000 ---
 TF_MODEL_NAME = "Anwarkh1/Skin_Cancer-Image_Classification"
@@ -25,6 +27,17 @@ model_vit.eval()
 model_malignancy = load_learner("ada_learn_malben.pkl")
 model_norm2000 = load_learner("ada_learn_skin_norm2000.pkl")
 
+# 🔹 EfficientNet B7 para binario (benigno vs maligno)
+model_eff = EfficientNet.from_pretrained("efficientnet-b7", num_classes=2)
+model_eff.eval()
+
+eff_transform = transforms.Compose([
+    transforms.Resize((224, 224)),
+    transforms.ToTensor(),
+    transforms.Normalize([0.485, 0.456, 0.406],
+                         [0.229, 0.224, 0.225])
+])
+
 # Clases estándar de HAM10000
 CLASSES = [
     "Queratosis actínica / Bowen", "Carcinoma células basales",
@@ -59,7 +72,7 @@ def analizar_lesion_combined(img):
     prob_malign = float(probs_mal[1])
     pred_fast_type, _, _ = model_norm2000.predict(img_fastai)
 
-    # ViT pre-trained fine-tuned (último modelo recomendado)
+    # ViT fine‑tuned (último modelo recomendado)
     inputs_tf = feature_extractor_tf(img, return_tensors="pt")
     with torch.no_grad():
         outputs_tf = model_tf_vit(**inputs_tf)
@@ -69,6 +82,13 @@ def analizar_lesion_combined(img):
     conf_tf = probs_tf[idx_tf]
     mal_tf = "Maligno" if idx_tf in MALIGNANT_INDICES else "Benigno"
 
+    # EfficientNet B7
+    img_eff = eff_transform(img).unsqueeze(0)
+    with torch.no_grad():
+        out_eff = model_eff(img_eff)
+        prob_eff = torch.softmax(out_eff, dim=1)[0, 1].item()
+    eff_result = "Maligno" if prob_eff > 0.5 else "Benigno"
+
     # Gráfico ViT base
     colors = [RISK_LEVELS[i]['color'] for i in range(7)]
     fig, ax = plt.subplots(figsize=(8, 3))
@@ -84,6 +104,7 @@ def analizar_lesion_combined(img):
     plt.close(fig)
     html_chart = f'<img src="data:image/png;base64,{base64.b64encode(buf.getvalue()).decode()}" style="max-width:100%"/>'
 
+    # Generar informe
     informe = f"""
     <div style="font-family:sans-serif; max-width:800px; margin:auto">
     <h2>🧪 Diagnóstico por múltiples modelos de IA</h2>
@@ -93,19 +114,22 @@ def analizar_lesion_combined(img):
       <tr><td>🧬 Fast.ai (tipo)</td><td><b>{pred_fast_type}</b></td><td>N/A</td></tr>
       <tr><td>⚠️ Fast.ai (malignidad)</td><td><b>{'Maligno' if prob_malign > 0.5 else 'Benigno'}</b></td><td>{prob_malign:.1%}</td></tr>
       <tr><td>🌟 ViT fined‑tuned (HAM10000)</td><td><b>{mal_tf} ({class_tf_model})</b></td><td>{conf_tf:.1%}</td></tr>
+      <tr><td>🏥 EfficientNet B7 (binario)</td><td><b>{eff_result}</b></td><td>{prob_eff:.1%}</td></tr>
     </table><br>
     <b>🩺 Recomendación automática:</b><br>
     """
+
+    # Nivel de riesgo automático
     risk = sum(probs_vit[i] * RISK_LEVELS[i]['weight'] for i in range(7))
-    if prob_malign > 0.7 or risk > 0.6:
+    if prob_malign > 0.7 or risk > 0.6 or prob_eff > 0.7:
         informe += "🚨 <b>CRÍTICO</b> – Derivación urgente a oncología dermatológica"
-    elif prob_malign > 0.4 or risk > 0.4:
+    elif prob_malign > 0.4 or risk > 0.4 or prob_eff > 0.5:
         informe += "⚠️ <b>ALTO RIESGO</b> – Consulta con dermatólogo en 7 días"
     elif risk > 0.2:
         informe += "📋 <b>RIESGO MODERADO</b> – Evaluación programada en 2-4 semanas"
     else:
         informe += "✅ <b>BAJO RIESGO</b> – Seguimiento de rutina (3-6 meses)"
-    informe += "</div>"""
+    informe += "</div>"
 
     return informe, html_chart
 
@@ -113,8 +137,10 @@ demo = gr.Interface(
     fn=analizar_lesion_combined,
     inputs=gr.Image(type="pil"),
     outputs=[gr.HTML(label="Informe"), gr.HTML(label="Gráfico ViT base")],
-    title="Detector de Lesiones Cutáneas (ViT + Fast.ai)",
+    title="Detector de Lesiones Cutáneas (ViT + Fast.ai + EfficientNet)",
 )
+
 if __name__ == "__main__":
     demo.launch()
 
+