首页 > 分享 > TensorFlow2实战

TensorFlow2实战

萌宠菠菠乐园
2024-11-19 05:13

TensorFlow2实战-系列教程总目录

有任何问题欢迎在下面留言
本篇文章的代码运行界面均在Jupyter Notebook中进行
本篇文章配套的代码资源已经上传

猫狗识别1
数据增强
猫狗识别2------数据增强
猫狗识别3------迁移学习

1、猫狗识别任务

base_dir = './data/cats_and_dogs' train_dir = os.path.join(base_dir, 'train') validation_dir = os.path.join(base_dir, 'validation') train_cats_dir = os.path.join(train_dir, 'cats') train_dogs_dir = os.path.join(train_dir, 'dogs') validation_cats_dir = os.path.join(validation_dir, 'cats') validation_dogs_dir = os.path.join(validation_dir, 'dogs') 123456789

model = tf.keras.models.Sequential([ tf.keras.layers.Conv2D(32, (3,3), activation='relu', input_shape=(64, 64, 3)), tf.keras.layers.MaxPooling2D(2, 2), tf.keras.layers.Conv2D(64, (3,3), activation='relu'), tf.keras.layers.MaxPooling2D(2,2), tf.keras.layers.Conv2D(128, (3,3), activation='relu'), tf.keras.layers.MaxPooling2D(2,2), tf.keras.layers.Flatten(), tf.keras.layers.Dense(512, activation='relu'), tf.keras.layers.Dense(1, activation='sigmoid') ]) model.compile(loss='binary_crossentropy', optimizer=Adam(lr=1e-4), metrics=['acc'])

1234567891011121314151617181920

依次是导包、指定数据路径、构建模型、配置训练器等，这些都与前面TensorFlow2实战-系列教程3：猫狗识别1完全一致

2、数据增强

train_datagen = ImageDataGenerator( rescale=1./255, rotation_range=40, width_shift_range=0.2, height_shift_range=0.2, shear_range=0.2, zoom_range=0.2, horizontal_flip=True, fill_mode='nearest') test_datagen = ImageDataGenerator(rescale=1./255) train_generator = train_datagen.flow_from_directory( train_dir, target_size=(64, 64), batch_size=20, class_mode='binary') validation_generator = test_datagen.flow_from_directory( validation_dir, target_size=(64, 64), batch_size=20, class_mode='binary') history = model.fit_generator( train_generator, steps_per_epoch=100, # 2000 images = batch_size * steps epochs=100, validation_data=validation_generator, validation_steps=50, # 1000 images = batch_size * steps verbose=2)

12345678910111213141516171819202122232425262728293031

train_datagen：

这里将rescale重新缩放、旋转、平移变换、剪切变换、缩放、水平翻转、以临近方式填充等多种方式对训练数据进行数据增强shear_range=0.2 表示图像将在 -0.2 到 +0.2 弧度的范围内随机剪切

test_datagen：

验证数据，没有进行数据增强，这里只进行了归一化操作

train_generator：

从 train_dir 目录加载训练图像，并应用前面定义的数据增强target_size=(64, 64)：调整图像大小为 64x64 像素batch_size=20：每批次处理 20 张图像class_mode='binary'：因为是二分类任务。

validation_generator：

从 validation_dir 目录加载验证图像，只应用缩放

history：

（fit_generator 方法在 TensorFlow 2.2 之后已经被弃用，建议使用 fit 方法替代）开始训练validation_data=validation_generator：指定验证数据生成器verbose=2：用于控制训练过程中输出的详细程度

3、预测效果展示

import matplotlib.pyplot as plt acc = history.history['acc'] val_acc = history.history['val_acc'] loss = history.history['loss'] val_loss = history.history['val_loss'] epochs = range(len(acc)) plt.plot(epochs, acc, 'b', label='Training accuracy') plt.plot(epochs, val_acc, 'r', label='Validation accuracy') plt.title('Training and validation accuracy') plt.legend() plt.figure() plt.plot(epochs, loss, 'b', label='Training Loss') plt.plot(epochs, val_loss, 'r', label='Validation Loss') plt.title('Training and validation loss') plt.legend() plt.show()

123456789101112131415161718192021

在这里插入图片描述

很显然经过数据增强后的模型表现对比原本效果有显著提升

3、加入Dropout

Dropout就是指定比例，对这一层随机杀死一下神经元，这里我们只需要在构建网络的时候在全连接层加上一层Dropout就可以了：

import os import warnings warnings.filterwarnings("ignore") import tensorflow as tf from tensorflow.keras.optimizers import Adam from tensorflow.keras.preprocessing.image import ImageDataGenerator base_dir = './data/cats_and_dogs' train_dir = os.path.join(base_dir, 'train') validation_dir = os.path.join(base_dir, 'validation') train_cats_dir = os.path.join(train_dir, 'cats') train_dogs_dir = os.path.join(train_dir, 'dogs') validation_cats_dir = os.path.join(validation_dir, 'cats') validation_dogs_dir = os.path.join(validation_dir, 'dogs') model = tf.keras.models.Sequential([ tf.keras.layers.Conv2D(32, (3,3), activation='relu', input_shape=(64, 64, 3)), tf.keras.layers.MaxPooling2D(2, 2), tf.keras.layers.Conv2D(64, (3,3), activation='relu'), tf.keras.layers.MaxPooling2D(2,2), tf.keras.layers.Conv2D(128, (3,3), activation='relu'), tf.keras.layers.MaxPooling2D(2,2), tf.keras.layers.Flatten(), tf.keras.layers.Dense(512, activation='relu'), tf.keras.layers.Dropout(0.5), tf.keras.layers.Dense(1, activation='sigmoid') ]) model.compile(loss='binary_crossentropy', optimizer=Adam(lr=1e-4), metrics=['acc']) train_datagen = ImageDataGenerator( rescale=1./255, rotation_range=40, width_shift_range=0.2, height_shift_range=0.2, shear_range=0.2, zoom_range=0.2, horizontal_flip=True, fill_mode='nearest') test_datagen = ImageDataGenerator(rescale=1./255) train_generator = train_datagen.flow_from_directory( train_dir, target_size=(64, 64), batch_size=20, class_mode='binary') validation_generator = test_datagen.flow_from_directory( validation_dir, target_size=(64, 64), batch_size=20, class_mode='binary') history = model.fit_generator( train_generator, steps_per_epoch=100, # 2000 images = batch_size * steps epochs=100, validation_data=validation_generator, validation_steps=50, # 1000 images = batch_size * steps verbose=2)

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566

Epoch 100/100
100/100 - 3s - loss: 0.4145 - acc: 0.8145 - val_loss: 0.4269 - val_acc: 0.7830 - 3s/epoch - 33ms/step

在这里插入图片描述

这效果又提升了一点