文章/答案/技术大牛

发布

社区首页 >问答首页 >无意输出形状:生成性对抗网络

问无意输出形状:生成性对抗网络
EN

Stack Overflow用户

提问于 2021-05-19 10:40:20

回答 2查看 27关注 0票数 0

所以我正在做这个项目，这个项目生成水果的图像，到目前为止，我已经编写了大部分实际代码，但我遇到了这个非常奇怪的问题。我遵循了https://towardsdatascience.com/gan-by-example-using-keras-on-tensorflow-backend-1a6d515a60d0关于手写数字的教程，虽然我可以弄乱它，但我的代码给了我错误，不知何故，输出形状(28，28，1)仍然显示在生成器模型中，而不是我的(100,100，3)，有人能解释我的这个奇怪的问题吗？

下面是我的代码：

import numpy as np;
import time;
#import training data

from keras.models import Sequential;
from keras.layers import Dense, Activation, Flatten, Reshape;
from keras.layers import Conv2D, Conv2DTranspose, UpSampling2D;
from keras.layers import LeakyReLU, Dropout;
from keras.layers import BatchNormalization;
from keras.optimizers import Adam, RMSprop;

class GAN(object):
    def __init__(self, img_rows, img_cols, channel):

        self.img_rows = img_rows;
        self.img_cols = img_cols;
        self.channel = channel;
        self.D = None;
        self.G = None;
        self.AM = None;
        self.DM = None;

    
    def discriminator(self):
        if self.D:
            return self.D;
        self.D = Sequential();
        depth = 64;
        dropout = 0.4;

        input_shape = (self.img_rows, self.img_cols, self.channel);
        self.D.add(Conv2D(depth*1, 5, strides=2, input_shape=input_shape, padding='same'));
        self.D.add(LeakyReLU(alpha=0.2));
        self.D.add(Dropout(dropout));

        self.D.add(Conv2D(depth*2, 5, strides=2, padding='same'));
        self.D.add(LeakyReLU(alpha=0.2));
        self.D.add(Dropout(dropout));

        self.D.add(Conv2D(depth*4, 5, strides=2, padding='same'));
        self.D.add(LeakyReLU(alpha=0.2));
        self.D.add(Dropout(dropout));

        self.D.add(Conv2D(depth*8, 5, strides=1, padding='same'));
        self.D.add(LeakyReLU(alpha=0.2));
        self.D.add(Dropout(dropout));

        self.D.add(Flatten());
        self.D.add(Dense(1));
        self.D.add(Activation('sigmoid'));
        self.D.summary();
        return self.D;

    def generator(self):
        if self.G:
            return self.G;
        self.G = Sequential();
        dropout = 0.4;
        depth = 64+64+64+64;
        dim = 7;
        
        self.G.add(Dense(dim*dim*depth, input_dim=100));
        self.G.add(BatchNormalization(momentum=0.9));
        self.G.add(Activation('relu'));
        self.G.add(Reshape((dim, dim, depth)));
        self.G.add(Dropout(dropout));

        self.G.add(UpSampling2D());
        self.G.add(Conv2DTranspose(int(depth/2), 5, padding='same'));
        self.G.add(BatchNormalization(momentum=0.9));
        self.G.add(Activation('relu'));


        self.G.add(UpSampling2D());
        self.G.add(Conv2DTranspose(int(depth/4), 5, padding='same'));
        self.G.add(BatchNormalization(momentum=0.9));
        self.G.add(Activation('relu'));

        self.G.add(Conv2DTranspose(int(depth/8), 5, padding='same'));
        self.G.add(BatchNormalization(momentum=0.9));
        self.G.add(Activation('relu'));

        self.G.add(Conv2DTranspose(1, 5, padding='same'));
        self.G.add(Activation('sigmoid'));
        self.G.summary();
        return self.G;

    def discriminator_model(self):
        if self.DM:
            return self.DM;
        optimizer = RMSprop(lr=0.0002, decay=6e-8);
        self.DM = Sequential();
        self.DM.add(self.discriminator());
        self.DM.compile(loss='binary_crossentropy', optimizer=optimizer, metrics=['accuracy']);
        return self.DM;

    def adversarial_model(self):
        if self.AM:
            return self.AM;
        optimizer = RMSprop(lr=0.0001, decay=3e-8);
        self.AM = Sequential();
        self.AM.add(self.generator());
        self.AM.add(self.discriminator());
        self.AM.compile(loss='binary_crossentropy', optimizer=optimizer, metrics=['accuracy']);
        return self.AM;

class Fruit_GAN:
    def __init__(self):
        self.img_rows = 28;
        self.img_cols = 28;
        self.channel = 1;
        print('\nstarting\n');
        self.GAN = GAN(self.img_rows, self.img_cols, self.channel);
        print('\nGAN setup succeded\n');
        self.discriminator = self.GAN.discriminator_model();
        print('\ndiscriminator setup succeded\n');
        self.adversarial = self.GAN.adversarial_model();
        print('\nadversarial setup succeded\n');
        self.generator = self.GAN.generator();
        print('\ngenerator setup succeded\n');
        
    def Train(self, train_steps=2000, batch_size=256, save_interval=0):
        noise_input = None;
        if save_interval>0:
            noise_input = np.random.uniform(-1.0, 1.0, size=[16, 100]);
        for i in range(train_steps):

            # training images:
            images_train = 0
            # random noise for generator:
            noise = np.random.uniform(-1.0, 1.0, size=[batch_size, 100]);
            # generator generated fake images:
            images_fake = self.generator.predict(noise);
            # train data formating for discriminator:
            x = np.concatenate((images_train, images_fake));
            y = np.ones([2*batch_size, 1]);
            y[batch_size:, :] = 0;
            # training_discriminator:
            d_loss = self.discriminator.train_on_batch(x, y);


            # train data reformatting for generator:
            y = np.ones([batch_size, 1]);
            noise = np.random.uniform(-1.0, 1.0, size=[batch_size, 100]);
            a_loss = self.adversarial.train_on_batch(noise, y);

            log_msg = "%d: [D loss: %f, acc: %f]" % (i, d_loss[0], d_loss[1]);
            log_msg = "%s: [A loss: %f, acc: %f]" % (log_mesg, a_loss[0], a_loss[1]);
            print(log_msg);

            if save_interval > 0:
                if (i+1) % save_interval == 0:
                    # save data
                    pass;

class ElapsedTimer(object):
    def __init__(self):
        self.start_time = time.time()
    def elapsed(self,sec):
        if sec < 60:
            return str(sec) + " sec"
        elif sec < (60 * 60):
            return str(sec / 60) + " min"
        else:
            return str(sec / (60 * 60)) + " hr"
    def elapsed_time(self):
        print("Elapsed: %s " % self.elapsed(time.time() - self.start_time) );
        

if __name__ == '__main__':
    f = Fruit_GAN();

machine-learning

keras

python

tensorflow

回答 2

Stack Overflow用户

发布于 2021-05-19 11:09:38

这是因为您指定了生成器的形状，我随意更改了代码以修复您的输出形状：(100,100，3)

G = tf.keras.Sequential()
dropout = 0.4
depth = 64+64+64+64
# I've change this to 25 as its a multiple of 7 before (7 * 4 = 28)
# So to fix that i've changed it to 25 (25 * 4 = 100)
dim = 25 

# In: 100
# Out: dim x dim x depth
G.add(tf.keras.layers.Dense(dim*dim*depth, input_dim=100))
G.add(tf.keras.layers.BatchNormalization(momentum=0.9))
G.add(tf.keras.layers.Activation('relu'))
G.add(tf.keras.layers.Reshape((dim, dim, depth)))
G.add(tf.keras.layers.Dropout(dropout))

# In: dim x dim x depth
# Out: 2*dim x 2*dim x depth/2
G.add(tf.keras.layers.UpSampling2D()) # Your upsampling layer is multiple of 2 
G.add(tf.keras.layers.Conv2DTranspose(int(depth/2), 5, padding='same'))
G.add(tf.keras.layers.BatchNormalization(momentum=0.9))
G.add(tf.keras.layers.Activation('relu'))
G.add(tf.keras.layers.UpSampling2D())
G.add(tf.keras.layers.Conv2DTranspose(int(depth/4), 5, padding='same'))
G.add(tf.keras.layers.BatchNormalization(momentum=0.9))
G.add(tf.keras.layers.Activation('relu'))
G.add(tf.keras.layers.Conv2DTranspose(int(depth/8), 5, padding='same'))
G.add(tf.keras.layers.BatchNormalization(momentum=0.9))
G.add(tf.keras.layers.Activation('relu'))

# Out: 100 x 100 x 3 rgb image [0.0,1.0] per pix
G.add(tf.keras.layers.Conv2DTranspose(3, 5, padding='same')) # Previously 1 representing 1 channel change to suit your rgb channel
G.add(tf.keras.layers.Activation('sigmoid'))
G.summary()

'''Final layer shape
activation_14 (Activation)   (None, 100, 100, 3)       0         
================================================================='''

票数 0

Stack Overflow用户

发布于 2021-05-19 12:07:07

您需要修改生成器部分。

首先把你的眼睛放在代码行上：

然后，...and将您的手指放到行上：

..。然后开始调整生成器的调光(尝试不同的调光值，并检查生成器的输出形状)，这样就可以看到调光对生成器输出的影响。如果需要，也可以调整生成器的其他层。这样，当图像大小发生变化时，您就会发现需要调整生成器参数。

代码的调试版本，现在产生大小为(100,100,1)的输出是：

import numpy as np;
import time;
#Those two imports were missing to reproduce and debug the problem...added...
import tensorflow as tf
from tensorflow import keras
#import training data

from keras.models import Sequential;
from keras.layers import Dense, Activation, Flatten, Reshape;
from keras.layers import Conv2D, Conv2DTranspose, UpSampling2D;
from keras.layers import LeakyReLU, Dropout;
from keras.layers import BatchNormalization;
from keras.optimizers import Adam, RMSprop;

class GAN(object):
    def __init__(self, img_rows, img_cols, channel):

        self.img_rows = img_rows;
        self.img_cols = img_cols;
        self.channel = channel;
        self.D = None;
        self.G = None;
        self.AM = None;
        self.DM = None;


    def discriminator(self):
        if self.D:
            return self.D;
        self.D = Sequential();
        depth = 64;
        dropout = 0.4;

        input_shape = (self.img_rows, self.img_cols, self.channel);
        print("Input shape nayttaa talta: ",input_shape)
        self.D.add(Conv2D(depth*1, 5, strides=2, input_shape=input_shape, padding='same'));
        self.D.add(LeakyReLU(alpha=0.2));
        self.D.add(Dropout(dropout));

        self.D.add(Conv2D(depth*2, 5, strides=2, padding='same'));
        self.D.add(LeakyReLU(alpha=0.2));
        self.D.add(Dropout(dropout));

        self.D.add(Conv2D(depth*4, 5, strides=2, padding='same'));
        self.D.add(LeakyReLU(alpha=0.2));
        self.D.add(Dropout(dropout));

        self.D.add(Conv2D(depth*8, 5, strides=1, padding='same'));
        self.D.add(LeakyReLU(alpha=0.2));
        self.D.add(Dropout(dropout));

        self.D.add(Flatten());
        self.D.add(Dense(1));
        self.D.add(Activation('sigmoid'));
        print("Diskriminaattori:")
        self.D.summary();
        print("Let's test it, note the shape of test input...")
        testitulos=self.D(tf.random.uniform((1,100,100,1)))
        print(testitulos)
        return self.D;

    def generator(self):
        if self.G:
            return self.G;
        self.G = Sequential();
        dropout = 0.4;
        depth = 64+64+64+64;
        #Let's modify the dim...
        dim = 25 #7;

        self.G.add(Dense(dim*dim*depth, input_dim=100));
        self.G.add(BatchNormalization(momentum=0.9));
        self.G.add(Activation('relu'));
        self.G.add(Reshape((dim, dim, depth)));
        self.G.add(Dropout(dropout));

        self.G.add(UpSampling2D());
        self.G.add(Conv2DTranspose(int(depth/2), 5, padding='same'));
        self.G.add(BatchNormalization(momentum=0.9));
        self.G.add(Activation('relu'));

        self.G.add(UpSampling2D());
        self.G.add(Conv2DTranspose(int(depth/4), 5, padding='same'));
        self.G.add(BatchNormalization(momentum=0.9));
        self.G.add(Activation('relu'));

        self.G.add(Conv2DTranspose(int(depth/8), 5, padding='same'));
        self.G.add(BatchNormalization(momentum=0.9));
        self.G.add(Activation('relu'));

        self.G.add(Conv2DTranspose(1, 5, padding='same'));
        self.G.add(Activation('sigmoid'));
        print("Generaattori:")
        self.G.summary();
        print("Let's test it, note the shape of the output of the generator...")
        testitulos=self.G(tf.random.uniform((1,100)))
        print(testitulos.shape)
        return self.G;

    def discriminator_model(self):
        if self.DM:
            return self.DM;
        optimizer = RMSprop(lr=0.0002, decay=6e-8);
        self.DM = Sequential();
        self.DM.add(self.discriminator());
        self.DM.compile(loss='binary_crossentropy', optimizer=optimizer, metrics=['accuracy']);
        return self.DM;

    def adversarial_model(self):
        if self.AM:
            return self.AM;
        optimizer = RMSprop(lr=0.0001, decay=3e-8);
        self.AM = Sequential();
        self.AM.add(self.generator());
        self.AM.add(self.discriminator());
        self.AM.compile(loss='binary_crossentropy', optimizer=optimizer, metrics=['accuracy']);
        return self.AM;

class Fruit_GAN:
    def __init__(self):
        self.img_rows = 100;
        self.img_cols = 100;
        self.channel = 1;
        print('\nstarting\n');
        self.GAN = GAN(self.img_rows, self.img_cols, self.channel);
        print('\nGAN setup succeded\n');
        self.discriminator = self.GAN.discriminator_model();
        print('\ndiscriminator setup succeded\n');
        self.adversarial = self.GAN.adversarial_model();
        print('\nadversarial setup succeded\n');
        self.generator = self.GAN.generator();
        print('\ngenerator setup succeded\n');

    def Train(self, train_steps=2000, batch_size=256, save_interval=0):
        noise_input = None;
        if save_interval>0:
            noise_input = np.random.uniform(-1.0, 1.0, size=[16, 100]);
        for i in range(train_steps):

            # training images:
            images_train = 0
            # random noise for generator:
            noise = np.random.uniform(-1.0, 1.0, size=[batch_size, 100]);
            # generator generated fake images:
            images_fake = self.generator.predict(noise);
            # train data formating for discriminator:
            x = np.concatenate((images_train, images_fake));
            y = np.ones([2*batch_size, 1]);
            y[batch_size:, :] = 0;
            # training_discriminator:
            d_loss = self.discriminator.train_on_batch(x, y);


            # train data reformatting for generator:
            y = np.ones([batch_size, 1]);
            noise = np.random.uniform(-1.0, 1.0, size=[batch_size, 100]);
            a_loss = self.adversarial.train_on_batch(noise, y);

            log_msg = "%d: [D loss: %f, acc: %f]" % (i, d_loss[0], d_loss[1]);
            log_msg = "%s: [A loss: %f, acc: %f]" % (log_mesg, a_loss[0], a_loss[1]);
            print(log_msg);

            if save_interval > 0:
                if (i+1) % save_interval == 0:
                    # save data
                    pass;

class ElapsedTimer(object):
    def __init__(self):
        self.start_time = time.time()
    def elapsed(self,sec):
        if sec < 60:
            return str(sec) + " sec"
        elif sec < (60 * 60):
            return str(sec / 60) + " min"
        else:
            return str(sec / (60 * 60)) + " hr"
    def elapsed_time(self):
        print("Elapsed: %s " % self.elapsed(time.time() - self.start_time) );


if __name__ == '__main__':
    f = Fruit_GAN();

票数 0

页面原文内容由Stack Overflow提供。腾讯云小微IT领域专用引擎提供翻译支持

原文链接：

https://stackoverflow.com/questions/67596192

复制

相似问题

问无意输出形状:生成性对抗网络
EN

回答 2

Stack Overflow用户

Stack Overflow用户

社区

活动

圈层

关于

腾讯云开发者

热门产品

热门推荐

更多推荐

问无意输出形状:生成性对抗网络EN

回答 2

Stack Overflow用户

Stack Overflow用户

社区

活动

圈层

关于

腾讯云开发者

热门产品

热门推荐

更多推荐

问无意输出形状:生成性对抗网络
EN