import tensorflow as tf
from face_recognition import config
import cv2
from glob import glob
import os
import numpy as np

feature_extractor=tf.keras.models.load_model("face_recognition/feature_extractor.h5",compile=False)
# feature_extractor.summary()

extensions=['.jpg','.jpeg','.png','.svg','.webp']

db_dir=config.db_dir
_,sub_folders,_=next(os.walk(db_dir))
print(sub_folders)
for sub_folder in sub_folders:
    image_paths=[]
    [image_paths.extend(glob(db_dir+"\\"+sub_folder+"\\*"+extension)) for extension in extensions]
    
    all_img_features=[]
    for image_path in image_paths:
        print(image_path)
        img=cv2.resize(cv2.imread(image_path),[config.input_size,config.input_size])
        all_img_features.append(feature_extractor.predict(img[None,:,:,::-1],verbose=0)[0])
    
    np.savetxt(db_dir+"\\"+sub_folder+"\\features.npy",all_img_features)
    
# "aligned_all"