GAN は漫画の顔を生成します

最近、私は対立生成ネットワークGANにもっと興味を持っており、関連する知識ポイントはまだ編集中です.以下は練習用の小さなプロジェクトです〜

 (計算量を減らし、トレーニングを容易にするために、元のモデルで行われました。)

1. ツールキットのインポート

import tensorflow as tf
from tensorflow.keras import layers

import numpy as np
import os
import time
import glob
import matplotlib.pyplot as plt
from IPython.display import clear_output
from IPython import display

1.1 GPU の設定

gpus = tf.config.list_physical_devices("GPU")

if gpus:
    gpu0 = gpus[0]                                        #如果有多个GPU，仅使用第0个GPU
    tf.config.experimental.set_memory_growth(gpu0, True)  #设置GPU显存用量按需使用
    tf.config.set_visible_devices([gpu0],"GPU")
gpus 

[PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')]

2.トレーニングデータをインポートする

リンク： ここをクリック

fileList = glob.glob('./ani_face/*.jpg')
len(fileList)

41621

2.1 データの視覚化 

# 随机显示几张图
for index,i in enumerate(fileList[:3]):
    display.display(display.Image(fileList[index]))

2.2 データの前処理

# 文件名列表
path_ds = tf.data.Dataset.from_tensor_slices(fileList)

# 预处理，归一化，缩放
def load_and_preprocess_image(path):
    image = tf.io.read_file(path)
    image = tf.image.decode_jpeg(image, channels=3)
    image = tf.image.resize(image, [64, 64])
    image /= 255.0  # normalize to [0,1] range
    image = tf.reshape(image, [1, 64,64,3])
    return image

image_ds = path_ds.map(load_and_preprocess_image)
image_ds

<MapDataset 形状: (1, 64, 64, 3), タイプ: tf.float32>

# 查看一张图片
for x in image_ds:
    plt.axis("off")
    plt.imshow((x.numpy() * 255).astype("int32")[0])
    break

3. ネットワーク構築

3.1D ネットワーク

discriminator = keras.Sequential(
    [
        keras.Input(shape=(64, 64, 3)),
        layers.Conv2D(64, kernel_size=4, strides=2, padding="same"),
        layers.LeakyReLU(alpha=0.2),
        layers.Conv2D(128, kernel_size=4, strides=2, padding="same"),
        layers.LeakyReLU(alpha=0.2),
        layers.Conv2D(128, kernel_size=4, strides=2, padding="same"),
        layers.LeakyReLU(alpha=0.2),
        layers.Flatten(),
        layers.Dropout(0.2),
        layers.Dense(1, activation="sigmoid"),
    ],
    name="discriminator",
)
discriminator.summary()

モデル: 「ディスクリミネーター」
_________________________________________________________________
レイヤー (タイプ) 出力シェイプ パラメータ #    
================================== ============================= 
conv2d (Conv2D) (なし、32、32、64) 3136       
__________________________________________________________________leaky_re_lu 
(LeakyReLU) (なし、32、32、64) 0          
_________________________________________________________________ conv2d_1 ( 
Conv2D 
) (なし、16、16、128)     
131200 
(LeakyReLU) (なし, 16, 16, 128) 0         
conv2d_2 (Conv2D) (なし、8、8、128) 262272     
_________________________________________________________________leaky_re_lu_2 
(LeakyReLU) (なし、8、8、128) 0          
_________________________________________________________________ 
flatten (平坦化) (なし、8192) 0          
_________________________________________________________________
ドロップアウト (ドロップアウト) (なし、8192) 0          
_________________________________________________________________
密 (密) (なし、1) 8193      
================================================== ===============
合計
パラメーター: 404,801トレーニング可能なパラメーター: 404,801 トレーニング不可能なパラメーター
: 0

3.2G ネットワーク

latent_dim = 128

generator = keras.Sequential(
    [
        keras.Input(shape=(latent_dim,)),
        layers.Dense(8 * 8 * 128),
        layers.Reshape((8, 8, 128)),
        layers.Conv2DTranspose(128, kernel_size=4, strides=2, padding="same"),
        layers.LeakyReLU(alpha=0.2),
        layers.Conv2DTranspose(256, kernel_size=4, strides=2, padding="same"),
        layers.LeakyReLU(alpha=0.2),
        layers.Conv2DTranspose(512, kernel_size=4, strides=2, padding="same"),
        layers.LeakyReLU(alpha=0.2),
        layers.Conv2D(3, kernel_size=5, padding="same", activation="sigmoid"),
    ],
    name="generator",
)
generator.summary()

3.3 書き換え train_step

class GAN(keras.Model):
    def __init__(self, discriminator, generator, latent_dim):
        super(GAN, self).__init__()
        self.discriminator = discriminator
        self.generator = generator
        self.latent_dim = latent_dim

    def compile(self, d_optimizer, g_optimizer, loss_fn):
        super(GAN, self).compile()
        self.d_optimizer = d_optimizer
        self.g_optimizer = g_optimizer
        self.loss_fn = loss_fn
        self.d_loss_metric = keras.metrics.Mean(name="d_loss")
        self.g_loss_metric = keras.metrics.Mean(name="g_loss")

    @property
    def metrics(self):
        return [self.d_loss_metric, self.g_loss_metric]

    def train_step(self, real_images):
        # 生成噪音
        batch_size = tf.shape(real_images)[0]
        random_latent_vectors = tf.random.normal(shape=(batch_size, self.latent_dim))

        # 生成的图片
        generated_images = self.generator(random_latent_vectors)

        # Combine them with real images
        combined_images = tf.concat([generated_images, real_images], axis=0)

        # Assemble labels discriminating real from fake images
        labels = tf.concat(
            [tf.ones((batch_size, 1)), tf.zeros((batch_size, 1))], axis=0
        )
        # Add random noise to the labels - important trick!
        labels += 0.05 * tf.random.uniform(tf.shape(labels))

        # 训练判别器，生成的当成0，真实的当成1 
        with tf.GradientTape() as tape:
            predictions = self.discriminator(combined_images)
            d_loss = self.loss_fn(labels, predictions)
        grads = tape.gradient(d_loss, self.discriminator.trainable_weights)
        self.d_optimizer.apply_gradients(
            zip(grads, self.discriminator.trainable_weights)
        )

        # Sample random points in the latent space
        random_latent_vectors = tf.random.normal(shape=(batch_size, self.latent_dim))

        # Assemble labels that say "all real images"
        misleading_labels = tf.zeros((batch_size, 1))

        # Train the generator (note that we should *not* update the weights
        # of the discriminator)!
        with tf.GradientTape() as tape:
            predictions = self.discriminator(self.generator(random_latent_vectors))
            g_loss = self.loss_fn(misleading_labels, predictions)
        grads = tape.gradient(g_loss, self.generator.trainable_weights)
        self.g_optimizer.apply_gradients(zip(grads, self.generator.trainable_weights))

        # Update metrics
        self.d_loss_metric.update_state(d_loss)
        self.g_loss_metric.update_state(g_loss)
        return {
            "d_loss": self.d_loss_metric.result(),
            "g_loss": self.g_loss_metric.result(),
        }

3.4 コールバック機能の設定

class GANMonitor(keras.callbacks.Callback):
    def __init__(self, num_img=3, latent_dim=128):
        self.num_img = num_img
        self.latent_dim = latent_dim

    def on_epoch_end(self, epoch, logs=None):
        random_latent_vectors = tf.random.normal(shape=(self.num_img, self.latent_dim))
        generated_images = self.model.generator(random_latent_vectors)
        generated_images *= 255
        generated_images.numpy()
        for i in range(self.num_img):
            img = keras.preprocessing.image.array_to_img(generated_images[i])
            display.display(img)
            img.save("gen_ani/generated_img_%03d_%d.png" % (epoch, i))

4. トレーニング モデル

epochs = 100  # In practice, use ~100 epochs

gan = GAN(discriminator=discriminator, generator=generator, latent_dim=latent_dim)
gan.compile(
    d_optimizer=keras.optimizers.Adam(learning_rate=0.0001),
    g_optimizer=keras.optimizers.Adam(learning_rate=0.0001),
    loss_fn=keras.losses.BinaryCrossentropy(),
)

gan.fit(
    image_ds, epochs=epochs, callbacks=[GANMonitor(num_img=10, latent_dim=latent_dim)]
)

5. モデルを保存する

#保存模型
gan.generator.save('./data/ani_G_model')

モデル ファイルの生成:ここをクリック

6.コミックフェイスを生成する

G_model =  tf.keras.models.load_model('./data/ani_G_model/',compile=False)

def randomGenerate():
    noise_seed = tf.random.normal([16, 128])
    predictions = G_model(noise_seed, training=False)
    fig = plt.figure(figsize=(8, 8))
    for i in range(predictions.shape[0]):
        plt.subplot(4, 4, i+1)
        img = (predictions[i].numpy() * 255 ).astype('int')
        plt.imshow(img )
        plt.axis('off')
    plt.show()

count = 0
while True:
    randomGenerate()
    clear_output(wait=True)
    time.sleep(0.1)
    if count > 100:
        break
    count+=1