ajout version cache heatmap

.gitignore

@@ -17,3 +17,6 @@ app/__pycache__/
 
 # Fichiers de logs
 *.log
+
+#cache des heatmaps
+ cache_heatmaps/

app/main.py

@@ -14,7 +14,6 @@ from app.log import logger
 from app.config import MODEL_NAME, ENV,MODEL_TYPE
 
 
-
 logger.info(f"ENV :{ENV}")
 
 app = FastAPI()

app/model.py

@@ -9,7 +9,7 @@ from keras.applications.efficientnet_v2 import preprocess_input as effnet_prepro
 import io
 from tf_keras_vis.gradcam import Gradcam,GradcamPlusPlus
 from tf_keras_vis.utils import normalize
-
+import hashlib
 import numpy as np
 import tensorflow as tf
 from tf_keras_vis.saliency import Saliency
@@ -20,8 +20,15 @@ from tf_keras_vis.saliency import Saliency
 from tf_keras_vis.utils import normalize
 import logging
 import time
+import os
 
 from typing import TypedDict, Callable, Any
+
+HEATMAP_CACHE = {}
+CACHE_DIR = "./cache_heatmaps"
+os.makedirs(CACHE_DIR, exist_ok=True)
+
+
 logging.basicConfig(
     level=logging.INFO,  # ou logging.DEBUG
     format="%(asctime)s [%(levelname)s] %(name)s: %(message)s"
@@ -322,7 +329,70 @@ def compute_entropy_safe(probas):
     return entropy
 
 
-def get_heatmap(config, image_bytes: bytes,predicted_class_index):
+def hash_image_bytes(image_bytes):
+    return hashlib.md5(image_bytes).hexdigest()
+
+
+def get_heatmap(config, image_bytes: bytes, predicted_class_index):
+    result = {}
+    try:
+        image_hash = hash_image_bytes(image_bytes)
+        cache_key = f"{image_hash}_{predicted_class_index}"
+
+        # Vérification cache mémoire d'abord
+        if cache_key in HEATMAP_CACHE:
+            logger.info(f"✅ Heatmap trouvée dans cache mémoire pour {cache_key}")
+            result["heatmap"] = HEATMAP_CACHE[cache_key]
+            return result
+
+        # Vérification cache disque ensuite
+        cache_file_path = os.path.join(CACHE_DIR, f"{cache_key}.pkl")
+        if os.path.exists(cache_file_path):
+            logger.info(f"✅ Heatmap trouvée sur disque pour {cache_key}")
+            with open(cache_file_path, "rb") as f:
+                cached_heatmap = pickle.load(f)
+            result["heatmap"] = cached_heatmap
+            # On remet aussi en mémoire pour accélérer prochaines requêtes
+            HEATMAP_CACHE[cache_key] = cached_heatmap
+            return result
+
+        # Calcul si non trouvé dans le cache
+        _, raw_input = preprocess_image(
+            image_bytes, config["target_size"], config["preprocess_input"]
+        )
+        logger.info("✅ Début de la génération de la heatmap")
+        start_time = time.time()
+
+        heatmap = compute_gradcam(
+            config["gradcam_model"],
+            raw_input,
+            class_index=predicted_class_index,
+            layer_name=config["last_conv_layer"],
+            gradcam_type=config["gradcam_type"],
+        )
+
+        elapsed_time = time.time() - start_time
+        logger.info(f"✅ Heatmap générée en {elapsed_time:.2f} secondes")
+
+        # Conversion en liste pour le JSON
+        heatmap_list = heatmap.tolist()
+        result["heatmap"] = heatmap_list
+
+        # Sauvegarde dans cache mémoire
+        HEATMAP_CACHE[cache_key] = heatmap_list
+
+        # Sauvegarde sur disque
+        with open(cache_file_path, "wb") as f:
+            pickle.dump(heatmap_list, f)
+
+    except Exception as e:
+        logger.error(f"❌ Erreur lors de la génération de la heatmap: {e}")
+        result["heatmap"] = []
+
+    return result
+
+
+def get_heatmap_old(config, image_bytes: bytes,predicted_class_index):
     result={}
     try:
         _,raw_input = preprocess_image(image_bytes,config["target_size"],config["preprocess_input"])

app/utils.py

@@ -30,7 +30,7 @@ def register_with_orchestrator():
         logger.info(f"📡 Tentative d'enregistrement de {MODEL_NAME} à l'orchestrateur...")
         response = requests.post(
             f"{ORCHESTRATOR_URL}/register_model",
-            json={"model_name": MODEL_NAME, "model_type": MODEL_TYPE,"url": f"{OWN_URL}/predict"}
+            json={"model_name": MODEL_NAME, "model_type": MODEL_TYPE,"url": f"{OWN_URL}"}
         )
         if response.status_code == 200:
             logger.info("✅ Modèle enregistré avec succès")