1. import numpy as np
    2. import cv2
    3. import os
    4. from sklearn.model_selection import train_test_split
    6. def open_img(path):
    7.     imgList = []    #图片
    8.     labels = []  #各图片对应的表情
    9.     label = []  #仅表情(无重复)
    10.     for file in os.listdir(path):
    11.         if file == '.anonr' or file == '.htaccess':
    12.             continue
    13.         position = path+'/'+file
    14.         for parent, dirnames, filenames in os.walk(position):
    15.             for filename in filenames:
    16.                 if filename == '.htaccess':
    17.                     continue
    18.                 fileList = filename.split('_')
    19.                 if fileList[2] not in label:
    20.                     label.append(fileList[2])
    21.                 labels.append(label.index(fileList[2]))
    22.                 img = cv2.imread(position+'/'+filename,cv2.IMREAD_GRAYSCALE)
    23.                 img = cv2.resize(img, (250, 350))
    24.                 imgList.append(img)
    25.     return imgList,labels
    27. def train_and_test(imgList,labels):
    28.     x_train,x_test,y_train,y_test = train_test_split(imgList,np.array(labels),test_size=0.3,random_state=42,stratify=np.array(labels))
    29.     recognizer = cv2.face.FisherFaceRecognizer_create()
    30.     recognizer.train(x_train,y_train)
    31.     right = 0
    32.     false = 0
    33.     for cnt in range(len(y_test)):
    34.         test_label = recognizer.predict(x_test[cnt])
    35.         if((test_label == y_test[cnt]).any()):
    36.             right += 1
    37.         else:
    38.             false += 1
    39.     return right/(right+false)
    41. if __name__ == '__main__':
    42.     path = "../faces"
    43.     imgList,labels = open_img(path)
    44.     print('表情识别的准确率为：',train_and_test(imgList,labels))