Tensorflow2.0 人脸识别

Tensorflow2.0使用sklearn内置的数据集进行人脸识别

首先准备数据集的下载，由于数据集是从国外的网站上下载，可能会报错

此时可以用此方法解决

import ssl

ssl._create_default_https_context = ssl._create_unverified_context

from sklearn import datasets

import tensorflow as tf

import matplotlib.pyplot as plt

from sklearn.model_selection import train_test_split

获取数据

faces = datasets.fetch_olivetti_faces()

测试获取的数据

plt.figure(figsize=(20, 25))

for index, img in enumerate(faces.images):

    plt.subplot(20, 20, index + 1)

    plt.imshow(img, cmap='gray')

#    关闭x轴

    plt.xticks([])

#    关闭y轴

    plt.yticks([])

    plt.xlabel(faces.target[index])

plt.show()

获取训练数据以及测试数据

X = faces.images

y = faces.target

X = X.reshape(400, 64, 64, 1)

抽取训练、测试数据

train_x, test_x, train_y, test_y = train_test_split(X, y, test_size=0.2)

建模

model = tf.keras.Sequential()

model.add(tf.keras.layers.Conv2D(128, kernel_size=3, activation='relu', input_shape=X.shape[1:]))

model.add(tf.keras.layers.Conv2D(64, kernel_size=3, activation='relu'))

model.add(tf.keras.layers.Flatten())

model.add(tf.keras.layers.Dense(40, activation='softmax'))

编译

model.compile(optimizer='adam',

              loss='sparse_categorical_crossentropy',

              metrics=['accuracy'])

开始训练

model.fit(train_x, train_y, epochs=8, validation_data=(test_x, test_y))

320/320 [==============================] - 7s 21ms/sample - loss: 5.0002 - accuracy: 0.0250 - val_loss: 3.6919 - val_accuracy: 0.0125

Epoch 2/9

320/320 [==============================] - 6s 19ms/sample - loss: 3.6688 - accuracy: 0.0656 - val_loss: 3.6290 - val_accuracy: 0.1250

Epoch 3/9

320/320 [==============================] - 6s 20ms/sample - loss: 3.4579 - accuracy: 0.1813 - val_loss: 3.5491 - val_accuracy: 0.0750

Epoch 4/9

320/320 [==============================] - 6s 20ms/sample - loss: 2.8424 - accuracy: 0.4563 - val_loss: 2.4117 - val_accuracy: 0.6000

Epoch 5/9

320/320 [==============================] - 7s 21ms/sample - loss: 1.5031 - accuracy: 0.7969 - val_loss: 1.5258 - val_accuracy: 0.7000

Epoch 6/9

320/320 [==============================] - 7s 22ms/sample - loss: 0.5492 - accuracy: 0.9187 - val_loss: 0.6792 - val_accuracy: 0.8500

Epoch 7/9

320/320 [==============================] - 7s 22ms/sample - loss: 0.1736 - accuracy: 0.9781 - val_loss: 0.7218 - val_accuracy: 0.7625

Epoch 8/9

320/320 [==============================] - 7s 21ms/sample - loss: 0.0640 - accuracy: 0.9969 - val_loss: 0.6178 - val_accuracy: 0.7875

当在第9个时候回过拟合