app.py

import os
import cv2
import dlib
import numpy as np
from PIL import Image
import gradio as gr
import git

# === Clone le dépôt GitHub contenant Known_faces_clean ===
REPO_URL = "https://github.com/Moustapha224/Reconnaissance"
REPO_DIR = "Reconnaissance"
KNOWN_FACES_DIR = os.path.join(REPO_DIR, "Known_faces_clean")

if not os.path.exists(REPO_DIR):
    print(f"🔄 Clonage du dépôt GitHub depuis {REPO_URL} ...")
    git.Repo.clone_from(REPO_URL, REPO_DIR)

# === Initialisation des outils Dlib ===
face_detector = dlib.get_frontal_face_detector()
shape_predictor = dlib.shape_predictor("shape_predictor_68_face_landmarks.dat")
face_rec_model = dlib.face_recognition_model_v1("dlib_face_recognition_resnet_model_v1.dat")

# === Stockage des visages connus ===
known_face_encodings = []
known_face_names = []

# === Fonction d'encodage d'une image en vecteur de visage ===
def get_face_encoding(image):
    if image is None or len(image.shape) != 3:
        raise ValueError("Image invalide (None ou pas RGB)")
    if image.shape[2] == 4:
        image = cv2.cvtColor(image, cv2.COLOR_RGBA2RGB)
    elif image.shape[2] != 3:
        image = cv2.cvtColor(image, cv2.COLOR_GRAY2RGB)

    dets = face_detector(image, 1)
    if len(dets) == 0:
        return None

    shape = shape_predictor(image, dets[0])
    face_descriptor = face_rec_model.compute_face_descriptor(image, shape)
    return np.array(face_descriptor)

# === Chargement des visages connus ===
def load_known_faces():
    print("🔍 Encodage des visages connus...")
    for person in os.listdir(KNOWN_FACES_DIR):
        person_path = os.path.join(KNOWN_FACES_DIR, person)
        if not os.path.isdir(person_path):
            continue
        for file in os.listdir(person_path):
            file_path = os.path.join(person_path, file)
            try:
                img = Image.open(file_path).convert("RGB")
                rgb = np.array(img)
                encoding = get_face_encoding(rgb)
                if encoding is not None:
                    known_face_encodings.append(encoding)
                    known_face_names.append(person)
                else:
                    print(f"⚠️ Aucun visage détecté dans {file_path}")
            except Exception as e:
                print(f"❌ Erreur avec {file_path}: {e}")

load_known_faces()

# === Fonction de reconnaissance faciale ===
def recognize_faces(image):
    if image is None:
        return None
    try:
        img = cv2.cvtColor(np.array(image), cv2.COLOR_RGBA2RGB if image.mode == "RGBA" else cv2.COLOR_RGB2BGR)
    except:
        return None

    dets = face_detector(img, 1)
    if len(dets) == 0:
        return image

    for d in dets:
        shape = shape_predictor(img, d)
        face_descriptor = face_rec_model.compute_face_descriptor(img, shape)
        encoding = np.array(face_descriptor)

        distances = np.linalg.norm(known_face_encodings - encoding, axis=1)
        min_distance = np.min(distances)
        name = "Inconnu"

        if min_distance < 0.6:
            name = known_face_names[np.argmin(distances)]

        # Dessiner le rectangle
        x1, y1, x2, y2 = d.left(), d.top(), d.right(), d.bottom()
        cv2.rectangle(img, (x1, y1), (x2, y2), (0, 255, 0), 2)
        cv2.putText(img, name, (x1, y2 + 20), cv2.FONT_HERSHEY_SIMPLEX, 0.8, (0, 255, 0), 2)

    return Image.fromarray(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))

# === Interface Gradio ===
iface = gr.Interface(
    fn=recognize_faces,
    inputs=gr.Image(label="Image à reconnaître"),
    outputs=gr.Image(label="Résultat"),
    title="Reconnaissance Faciale",
    description="Téléversez une image contenant un visage pour l'identifier à partir des visages connus."
)

iface.launch()