tf中的卷积神经网络实战

CIFAR 100实战

1. 思路（卷积层+池化层+…+卷积层+池化层+全连层+输出层）

• 加载数据集，数据集预处理
• 设定卷积层与池化层
• 设定全连接层（设定卷积层池化层与全连层的输入维度）
• 梯度下降，更新参数
• 测试（获得准确率）

2. 代码

import tensorflow as tf
from tensorflow.keras import layers, optimizers, datasets, Sequential
import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'


def pre_process(x, y):
    x = tf.cast(x, dtype=tf.float32) / 255.
    y = tf.cast(y, dtype=tf.int32)
    return x, y


batch_size = 129
(x, y), (x_test, y_test) = datasets.cifar100.load_data()
# 将标签降维
y = tf.squeeze(y, axis=1)
y_test = tf.squeeze(y_test, axis=1)
train_db = tf.data.Dataset.from_tensor_slices((x, y))
train_db = train_db.shuffle(1000).map(pre_process).batch(batch_size)
test_db = tf.data.Dataset.from_tensor_slices((x_test, y_test))
test_db = test_db.map(pre_process).batch(batch_size)
sample = next(iter(train_db))


conv_layers = [
    layers.Conv2D(64, kernel_size=[3, 3], padding='same', activation=tf.nn.relu),
    layers.Conv2D(64, kernel_size=[3, 3], padding='same', activation=tf.nn.relu),
    layers.MaxPool2D(pool_size=[2, 2], strides=2, padding='same'),
    layers.Conv2D(128, kernel_size=[3, 3], padding='same', activation=tf.nn.relu),
    layers.Conv2D(128, kernel_size=[3, 3], padding='same', activation=tf.nn.relu),
    layers.MaxPool2D(pool_size=[2, 2], strides=2, padding='same'),
    layers.Conv2D(256, kernel_size=[3, 3], padding='same', activation=tf.nn.relu),
    layers.Conv2D(256, kernel_size=[3, 3], padding='same', activation=tf.nn.relu),
    layers.MaxPool2D(pool_size=[2, 2], strides=2, padding='same'),
    layers.Conv2D(512, kernel_size=[3, 3], padding='same', activation=tf.nn.relu),
    layers.Conv2D(512, kernel_size=[3, 3], padding='same', activation=tf.nn.relu),
    layers.MaxPool2D(pool_size=[2, 2], strides=2, padding='same'),
    layers.Conv2D(512, kernel_size=[3, 3], padding='same', activation=tf.nn.relu),
    layers.Conv2D(512, kernel_size=[3, 3], padding='same', activation=tf.nn.relu),
    layers.MaxPool2D(pool_size=[2, 2], strides=2, padding='same')
]


def main():
    con_net = Sequential(conv_layers)
    fc_net = Sequential([
        layers.Dense(256, activation=tf.nn.relu),
        layers.Dense(128, activation=tf.nn.relu),
        layers.Dense(100, activation=tf.nn.relu)  # 因为是分100类
    ])
    # 控制卷积层与全连层的输入维度
    con_net.build(input_shape=[None, 32, 32, 3])
    fc_net.build(input_shape=[None, 512])
    optimizer = optimizers.Adam(lr=1e-4)
    # 将两组网络需要训练的参数连接到一起
    variables = con_net.trainable_variables + fc_net.trainable_variables
    for epoch in range(100):
        for step, (x, y) in enumerate(train_db):
            with tf.GradientTape() as tape:
                out = con_net(x)
                out = tf.reshape(out, [-1, 512])
                output = fc_net(out)
                y_onehot = tf.one_hot(y, depth=100)
                loss = tf.losses.categorical_crossentropy(y_onehot, output, from_logits=True)
                loss = tf.reduce_mean(loss)
            grads = tape.gradient(loss, variables)
            optimizer.apply_gradients(zip(grads, variables))
            if step % 100 == 0:
                print(epoch, step, 'loss:', float(loss))

        # 测试
        total_num = 0
        total_correct = 0
        for x,y in test_db:
            out = con_net(x)
            out = tf.reshape(out, [-1, 512])
            output = fc_net(out)
            prob = tf.nn.softmax(output, axis=1)
            pred = tf.argmax(prob, axis=1)
            pred = tf.cast(pred, dtype=tf.int32)
            correct = tf.cast(tf.equal(pred, y), dtype=tf.int32)
            correct = tf.reduce_sum(correct)
            total_num += x.shape[0]
            total_correct += int(correct)
        accuracy = total_correct / total_num
        accuracy = accuracy * 100
        print(epoch, '准确率', accuracy, "%")


if __name__ == '__main__':
    main()