ValueError：无法为张量u'InputData/X：0'，形状为'（？，32，32，1）'的形状（64,32,32）提供值'

Question

我试图使用tflearn和我自己的数据。

我有19748个灰度图像，我想用我的模型进行训练。我使用了tflearn的Image_Preloader方法来输入图像。所有图像都转换成32 * 32大小。但是当我开始训练过程时，我得到这个错误“ValueError：无法提供形状为'（？，32，32，1）的Tensor u'InputData/X：0'的形状值（64,32,32） '“

我已经尝试了一切在我的知识，但我不能解决它，并有类似的问题在stackoverflow中，但他们没有为我工作。

这是我的代码。

from __future__ import division, print_function, absolute_import 


import tflearn 
import pickle 
from tflearn.layers.core import input_data, dropout, fully_connected 
from tflearn.layers.conv import conv_2d, max_pool_2d 
from tflearn.layers.normalization import local_response_normalization 
from tflearn.layers.estimator import regression 
from time import gmtime, strftime 
from tflearn.data_utils import image_preloader 
import numpy as np 


dataset_file = 'noww.txt' 



X = np.zeros((19748,32,32,1)) 
Y = np.zeros((19748,10)) 

X, Y = image_preloader(dataset_file, image_shape=(32, 32), mode='file', categorical_labels=True, normalize=True) 


network = input_data(shape=[None, 32, 32, 1]) 


network = conv_2d(network, 64, 3, activation='relu') 
network = conv_2d(network, 64, 3, activation='relu') 
network = max_pool_2d(network, 2, strides=2) 

network = conv_2d(network, 128, 3, activation='relu') 
network = conv_2d(network, 128, 3, activation='relu') 
network = max_pool_2d(network, 2, strides=2) 

network = conv_2d(network, 256, 3, activation='relu') 
network = conv_2d(network, 256, 3, activation='relu') 
network = conv_2d(network, 256, 3, activation='relu') 
network = max_pool_2d(network, 2, strides=2) 

network = fully_connected(network, 1024, activation='relu') 
network = dropout(network, 0.5) 
network = fully_connected(network, 1024, activation='relu') 
network = dropout(network, 0.5) 
network = fully_connected(network, 10, activation='softmax') 
network = regression(network, optimizer='rmsprop', 
        loss='categorical_crossentropy', 
        learning_rate=0.0001) 


model = tflearn.DNN(network, checkpoint_path='model_1', 
        max_checkpoints=1, tensorboard_verbose=0) 
model.fit(X, Y, n_epoch=200, shuffle=True, 
      show_metric=True, batch_size=64, snapshot_step=200, 
      snapshot_epoch=False, run_id='model_1') 

请帮忙。

Answer 1

错误表明Tensorflow不能将具有形状[64,32,32]的张量放置到形状为[？，32,32,1]的另一个张量中，这里的?表示批量大小。

您的模型无法将批处理数据送入X变量，因为它们的形状不同，您应该更改X形状。

改变这一行X, Y = image_preloader(dataset_file, image_shape=(32, 32), mode='file', categorical_labels=True, normalize=True)

到

X, Y = image_preloader(dataset_file, image_shape=(None, 32, 32, 1), mode='file', categorical_labels=True, normalize=True)

希望那是有益的。