Я пытаюсь создать Variational Autoencoder, который должен быть обучен на изображениях некоторых кошек (224,224,3). После создания архитектуры (с помощью нескольких учебных пособий, так как я новичок в автоэнкодере, [основной источник] [1]), я попытался согласовать модель с имеющимися у меня данными. Но я получил ошибку
InvalidArgumentError: Несовместимые формы: [40,224,224] против [40] [[градиенты узла / add_1_grad / BroadcastGradientArgs (определены в / Users / Philipp / anaconda3 / lib / python3 6 ошибка, которая была бы велика. Заранее спасибо!
Система:
- macOS 10.15.2
- Python 3.6.9
- Керас 2.3.1
- TF 2.0.0
from ProcessData import process_dataset
import matplotlib.pyplot as plt
from keras.utils import plot_model
from keras.losses import binary_crossentropy
X_train,y_train,X_test,y_test = process_dataset(size=(224,224))
X_train_cats = X_train[:200]
X_test_cats = X_test[:40]
process_dataset возвращает X_train (1000,224,224,3) и X_test (200,224,224)
from keras.layers import Input, Dense, Conv2D, MaxPooling2D, UpSampling2D, Conv2DTranspose, Flatten, Reshape, GlobalAveragePooling2D,Lambda
from keras.models import Model,Sequential
import keras as K
import tensorflow as tf
number_training = len(X_train_cats)
batch_size = 40
epochs = 20
latent_space = 10
image_size = 224
Функция кодирования
def encoder():
input_img = Input(shape=(224, 224, 3),name="encoder_input") # adapt this if using `channels_first` image data format
x = Conv2D(32, (3, 3), activation='relu',padding="same")(input_img)
x = MaxPooling2D((2, 2))(x)
x = Conv2D(64, (3, 3), activation='relu',padding="same")(x)
x = MaxPooling2D((2, 2))(x)
x = Conv2D(128, (3, 3), activation='relu',padding="same")(x)
x = MaxPooling2D((2, 2))(x)
x = Flatten()(x)
z_mean = Dense(latent_space,name="z_mean")(x)
z_log_var = Dense(latent_space,name="z_std")(x)
z = Lambda(sampling, output_shape=(latent_space,))([z_mean, z_log_var])
encoder = Model(input_img, [z_mean, z_log_var, z], name='encoder')
return encoder, input_img, z_mean, z_log_var
Выборочная функция
def sampling(args):
"""Reparameterization trick by sampling from an isotropic unit Gaussian.
# Arguments
args (tensor): mean and log of variance of Q(z|X)
# Returns
z (tensor): sampled latent vector
"""
#dim should be correct (40,10)
z_mean, z_log_var = args
batch = K.backend.shape(z_mean)[0]
dim = K.backend.int_shape(z_mean)[1]
# by default, random_normal has mean = 0 and std = 1.0
epsilon = K.backend.random_normal(shape=(batch, dim))
return z_mean + K.backend.exp(0.5 * z_log_var) * epsilon
Декодер
def decoding():
latent_inputs = Input(shape=(latent_space,), name='z_sampling')
x = Dense(100352)(latent_inputs)
x = Reshape((28,28,128))(x)
x = Conv2D(128, (3, 3), activation='relu', padding='same')(x)
x = UpSampling2D((2, 2))(x)
x = Conv2D(64, (3, 3), activation='relu', padding='same')(x)
x = UpSampling2D((2, 2))(x)
x = Conv2D(32, (3, 3), activation='relu',padding="same")(x)
x = UpSampling2D((2, 2))(x)
outputs = Conv2DTranspose(3, (3, 3), activation='sigmoid',padding="same")(x)
# instantiate decoder model
decoder = Model(latent_inputs, outputs, name='decoder')
return decoder,outputs
Собираем все вместе
encoder, input_img, z_mean, z_log_var = encoder()
encoder.summary()
plot_model(encoder, to_file='vae_mlp_encoder.png', show_shapes=True,rankdir="LR",dpi=100)
decoder, outputs = decoding()
decoder.summary()
plot_model(decoder, to_file='vae_mlp_decoder.png', show_shapes=True,rankdir="LR",dpi=100)
# instantiate VAE model
outputs = decoder(encoder(input_img)[2])
vae = Model(input_img, outputs, name='vae_mlp')
reconstruction_loss = binary_crossentropy(input_img,outputs)
original_dim = image_size * image_size
reconstruction_loss *= original_dim
kl_loss = 1 + z_log_var - K.backend.square(z_mean) - K.backend.exp(z_log_var)
kl_loss = K.backend.sum(kl_loss, axis=-1)
kl_loss *= -0.5
vae_loss = K.backend.mean(reconstruction_loss + kl_loss)
vae.add_loss(vae_loss)
vae.compile(optimizer='adam')
vae.summary()
plot_model(vae,
to_file='vae_mlp.png',
show_shapes=True)
Пока здесь все работает
vae.fit(X_train_cats,
epochs=epochs,
batch_size=batch_size,
validation_data=(X_test_cats,None))
Ошибка
Train on 200 samples, validate on 40 samples
Epoch 1/20
---------------------------------------------------------------------------
InvalidArgumentError Traceback (most recent call last)
<ipython-input-15-e4eec5a6e821> in <module>
2 epochs=epochs,
3 batch_size=batch_size,
----> 4 validation_data=(X_test_cats,None))
~/anaconda3/lib/python3.6/site-packages/keras/engine/training.py in fit(self, x, y, batch_size, epochs, verbose, callbacks, validation_split, validation_data, shuffle, class_weight, sample_weight, initial_epoch, steps_per_epoch, validation_steps, validation_freq, max_queue_size, workers, use_multiprocessing, **kwargs)
1237 steps_per_epoch=steps_per_epoch,
1238 validation_steps=validation_steps,
-> 1239 validation_freq=validation_freq)
1240
1241 def evaluate(self,
~/anaconda3/lib/python3.6/site-packages/keras/engine/training_arrays.py in fit_loop(model, fit_function, fit_inputs, out_labels, batch_size, epochs, verbose, callbacks, val_function, val_inputs, shuffle, initial_epoch, steps_per_epoch, validation_steps, validation_freq)
194 ins_batch[i] = ins_batch[i].toarray()
195
--> 196 outs = fit_function(ins_batch)
197 outs = to_list(outs)
198 for l, o in zip(out_labels, outs):
~/anaconda3/lib/python3.6/site-packages/tensorflow_core/python/keras/backend.py in __call__(self, inputs)
3738 value = math_ops.cast(value, tensor.dtype)
3739 converted_inputs.append(value)
-> 3740 outputs = self._graph_fn(*converted_inputs)
3741
3742 # EagerTensor.numpy() will often make a copy to ensure memory safety.
~/anaconda3/lib/python3.6/site-packages/tensorflow_core/python/eager/function.py in __call__(self, *args, **kwargs)
1079 TypeError: For invalid positional/keyword argument combinations.
1080 """
-> 1081 return self._call_impl(args, kwargs)
1082
1083 def _call_impl(self, args, kwargs, cancellation_manager=None):
~/anaconda3/lib/python3.6/site-packages/tensorflow_core/python/eager/function.py in _call_impl(self, args, kwargs, cancellation_manager)
1119 raise TypeError("Keyword arguments {} unknown. Expected {}.".format(
1120 list(kwargs.keys()), list(self._arg_keywords)))
-> 1121 return self._call_flat(args, self.captured_inputs, cancellation_manager)
1122
1123 def _filtered_call(self, args, kwargs):
~/anaconda3/lib/python3.6/site-packages/tensorflow_core/python/eager/function.py in _call_flat(self, args, captured_inputs, cancellation_manager)
1222 if executing_eagerly:
1223 flat_outputs = forward_function.call(
-> 1224 ctx, args, cancellation_manager=cancellation_manager)
1225 else:
1226 gradient_name = self._delayed_rewrite_functions.register()
~/anaconda3/lib/python3.6/site-packages/tensorflow_core/python/eager/function.py in call(self, ctx, args, cancellation_manager)
509 inputs=args,
510 attrs=("executor_type", executor_type, "config_proto", config),
--> 511 ctx=ctx)
512 else:
513 outputs = execute.execute_with_cancellation(
~/anaconda3/lib/python3.6/site-packages/tensorflow_core/python/eager/execute.py in quick_execute(op_name, num_outputs, inputs, attrs, ctx, name)
65 else:
66 message = e.message
---> 67 six.raise_from(core._status_to_exception(e.code, message), None)
68 except TypeError as e:
69 keras_symbolic_tensors = [
~/anaconda3/lib/python3.6/site-packages/six.py in raise_from(value, from_value)
InvalidArgumentError: Incompatible shapes: [40,224,224] vs. [40]
[[node add_1 (defined at /Users/Philipp/anaconda3/lib/python3.6/site-packages/tensorflow_core/python/framework/ops.py:1751) ]] [Op:__inference_keras_scratch_graph_3165]
Function call stack:
keras_scratch_graph
[1]: https://github.com/keras-team/keras/blob/master/examples/variational_autoencoder.py