import os # Utilise un dossier temporaire où on a le droit d'écriture deepface_home = "/tmp/.deepface" os.environ['DEEPFACE_HOME'] = deepface_home os.makedirs(deepface_home, exist_ok=True) import json import numpy as np from deepface import DeepFace import cv2 def auto_rotate_if_needed(image_path): img = cv2.imread(image_path) if img is None: raise ValueError(f"Image non lisible : {image_path}") h, w = img.shape[:2] # 1. Si image couchée → pivoter verticalement if w > h: img = cv2.rotate(img, cv2.ROTATE_90_CLOCKWISE) # 2. Si image à l’envers (tête en bas), corriger : # On peut ici forcer une rotation supplémentaire si nécessaire # 💡 OPTION : ajouter une vérification automatique plus intelligente avec un modèle plus tard img = cv2.rotate(img, cv2.ROTATE_180) # Pour le cas à l’envers # 3. Écraser l'image modifiée cv2.imwrite(image_path, img) return image_path def auto_rotate_folder(folder_path): for filename in os.listdir(folder_path): if filename.lower().endswith((".jpg", ".jpeg", ".png")): full_path = os.path.join(folder_path, filename) auto_rotate_if_needed(full_path) def auto_rotate_all(folders): """ Corrige l'orientation des images dans une liste de dossiers, y compris leurs sous-dossiers. """ for folder_path in folders: print(f"\n🔁 Correction des images dans : {folder_path}") for root, _, files in os.walk(folder_path): for filename in files: if filename.lower().endswith((".jpg", ".jpeg", ".png")): full_path = os.path.join(root, filename) try: auto_rotate_if_needed(full_path) except Exception as e: print(f"⚠️ Erreur sur {filename} : {e}") def create_template_from_folder(folder_path, model_name="ArcFace"): embeddings = [] for filename in os.listdir(folder_path): if filename.lower().endswith((".jpg", ".jpeg", ".png")): img_path = os.path.join(folder_path, filename) try: print(f"📷 Traitement : {filename}") auto_rotate_if_needed(img_path) reps = DeepFace.represent(img_path=img_path, model_name=model_name, enforce_detection=True) embeddings.append(reps[0]["embedding"]) except Exception as e: print(f"⚠️ Erreur sur {filename} : {e}") if not embeddings: raise RuntimeError("❌ Aucun embedding valide extrait.") return np.mean(np.array(embeddings), axis=0) def save_template(template_array, output_path): with open(output_path, "w") as f: json.dump(template_array.tolist(), f) print(f"✅ Template sauvegardé dans : {output_path}") def load_template(input_path): with open(input_path, "r") as f: return np.array(json.load(f)) def verify_against_template(image_path, template_array, threshold=0.5, model_name="ArcFace"): reps = DeepFace.represent(img_path=image_path, model_name=model_name, enforce_detection=True) embedding = np.array(reps[0]["embedding"]) similarity = np.dot(template_array, embedding) / (np.linalg.norm(template_array) * np.linalg.norm(embedding)) return similarity >= threshold, similarity