CIFAR10與VGG13實戰

1. 準備資料

CIFAR10 資料集由加拿大 Canadian Institute For Advanced Research 發布，它包含了飛機、汽車、鳥、貓等共 10 大類物體的彩色圖片，每個種類收集了 6000 張 32x32 大小圖片，共 60K 張圖片，其中 50K 作為訓練資料集，10K 作為測驗資料集，

import tensorflow as tf
from tensorflow.keras import datasets,layers,losses,optimizers,Sequential
import  os

os.environ['TF_CPP_MIN_LOG_LEVEL']='2'

使用datasets.cifar10.load_data()加載資料集,并對資料集進行預處理

def preprocess(x, y):
    # [0~1]
    x = 2*tf.cast(x, dtype=tf.float32) / 255.-1
    y = tf.cast(y, dtype=tf.int32)
    return x,y


(x,y), (x_test, y_test) = datasets.cifar10.load_data()
y = tf.squeeze(y, axis=1)
y_test = tf.squeeze(y_test, axis=1)
print(x.shape, y.shape, x_test.shape, y_test.shape)

(50000, 32, 32, 3) (50000,) (10000, 32, 32, 3) (10000,)

train_db = tf.data.Dataset.from_tensor_slices((x,y))
train_db = train_db.shuffle(1000).map(preprocess).batch(128)

test_db = tf.data.Dataset.from_tensor_slices((x_test,y_test))
test_db = test_db.map(preprocess).batch(64)

sample = next(iter(train_db))
print('sample:', sample[0].shape, sample[1].shape,
      tf.reduce_min(sample[0]), tf.reduce_max(sample[0]))

sample: (128, 32, 32, 3) (128,) tf.Tensor(-1.0, shape=(), dtype=float32) tf.Tensor(1.0, shape=(), dtype=float32)

2.構建網路模型

CIFAR10 圖片識別任務并不簡單，這主要是由于 CIFAR10 的圖片內容需要大量細節才能呈現，而保存的圖片解析度僅有 32x32，使得部分主體資訊較為模糊，甚至人眼都很難分辨，淺層的神經網路表達能力有限，很難訓練優化到較好的性能，本節將基于表達能力更強的 VGG13 網路，根據我們的資料集特點修改部分網路結構，完成 CIFAR10 圖片識別，

我們將網路實作為 2 個子網路：卷積子網路和全連接子網路，卷積子網路由 5 個子模塊構成，每個子模塊包含了 Conv-Conv-MaxPooling 單元結構：

conv_layers = [ # 先創建包含多層的串列
    # unit 1
    # 64 個 3x3 卷積核, 輸入輸出同大小
    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'),

    # unit 2
    #輸出通道提升至 128，高寬大小減半
    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'),

    # unit 3
    #,輸出通道提升至 256，高寬大小減半
    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'),

    # unit 4
    #輸出通道提升至 512，高寬大小減半
    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'),

    # unit 5
    #輸出通道提升至 512，高寬大小減半
    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')
]
# 利用前面創建的層串列構建網路容器
conv_net = Sequential(conv_layers)

全連接子網路包含了 3 個全連接層，每層添加 ReLU 非線性激活函式，最后一層除外

# 創建 3 層全連接層子網路
fc_net = Sequential([
    layers.Dense(256, activation=tf.nn.relu),
    layers.Dense(128, activation=tf.nn.relu),
    layers.Dense(10, activation=None),
])
# build2 個子網路，并列印網路引數資訊
conv_net.build(input_shape=[4, 32, 32, 3])
fc_net.build(input_shape=[4, 512])
conv_net.summary()

Model: "sequential"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d (Conv2D)              (4, 32, 32, 64)           1792      
_________________________________________________________________
conv2d_1 (Conv2D)            (4, 32, 32, 64)           36928     
_________________________________________________________________
max_pooling2d (MaxPooling2D) (4, 16, 16, 64)           0         
_________________________________________________________________
conv2d_2 (Conv2D)            (4, 16, 16, 128)          73856     
_________________________________________________________________
conv2d_3 (Conv2D)            (4, 16, 16, 128)          147584    
_________________________________________________________________
max_pooling2d_1 (MaxPooling2 (4, 8, 8, 128)            0         
_________________________________________________________________
conv2d_4 (Conv2D)            (4, 8, 8, 256)            295168    
_________________________________________________________________
conv2d_5 (Conv2D)            (4, 8, 8, 256)            590080    
_________________________________________________________________
max_pooling2d_2 (MaxPooling2 (4, 4, 4, 256)            0         
_________________________________________________________________
conv2d_6 (Conv2D)            (4, 4, 4, 512)            1180160   
_________________________________________________________________
conv2d_7 (Conv2D)            (4, 4, 4, 512)            2359808   
_________________________________________________________________
max_pooling2d_3 (MaxPooling2 (4, 2, 2, 512)            0         
_________________________________________________________________
conv2d_8 (Conv2D)            (4, 2, 2, 512)            2359808   
_________________________________________________________________
conv2d_9 (Conv2D)            (4, 2, 2, 512)            2359808   
_________________________________________________________________
max_pooling2d_4 (MaxPooling2 (4, 1, 1, 512)            0         
=================================================================
Total params: 9,404,992
Trainable params: 9,404,992
Non-trainable params: 0
_________________________________________________________________

3.訓練模型

下面對資料進行訓練，并對測驗資料集進行測驗

def main():

    # [b, 32, 32, 3] => [b, 1, 1, 512]
    conv_net = Sequential(conv_layers)

    fc_net = Sequential([
        layers.Dense(256, activation=tf.nn.relu),
        layers.Dense(128, activation=tf.nn.relu),
        layers.Dense(10, activation=None),
    ])

    conv_net.build(input_shape=[None, 32, 32, 3])
    fc_net.build(input_shape=[None, 512])
#     conv_net.summary()
#     fc_net.summary()
    optimizer = optimizers.Adam(lr=1e-4)

    # [1, 2] + [3, 4] => [1, 2, 3, 4]
    variables = conv_net.trainable_variables + fc_net.trainable_variables

    for epoch in range(50):

        for step, (x,y) in enumerate(train_db):

            with tf.GradientTape() as tape:
                # [b, 32, 32, 3] => [b, 1, 1, 512]
                out = conv_net(x)
                # flatten, => [b, 512]
                out = tf.reshape(out, [-1, 512])
                # [b, 512] => [b, 10]
                logits = fc_net(out)
                # [b] => [b, 10]
                y_onehot = tf.one_hot(y, depth=10)
                # compute loss
                loss = tf.losses.categorical_crossentropy(y_onehot, logits, 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))



        total_num = 0
        total_correct = 0
        for x,y in test_db:

            out = conv_net(x)
            out = tf.reshape(out, [-1, 512])
            logits = fc_net(out)
            prob = tf.nn.softmax(logits, 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)

        acc = total_correct / total_num
        print(epoch, 'acc:', acc)



if __name__ == '__main__':
    main()