flow_from_directory 用于 one-hot 编码 python 错误

问题描述

我执行 class_mode='categorical' 并且我的文件夹名称是 0 到 1801 的整数，但是当我运行代码时它返回此错误：

ValueError：检查目标时出错：预期dense_4具有形状（无，1082）但得到形状为（16，1802）的数组

我的模型是：

def Create_Model(conv_base):

model = models.Sequential()
model.add(conv_base)
model.add(layers.Dropout(0.5))
model.add(layers.Flatten())
model.add(layers.Dense(14,activation='relu'))
# model.add(layers.Dense(1,activation='sigmoid'))
model.add(layers.Dropout(0.5))
model.add(layers.Dense(1082,activation='softmax'))

# Compile Model
model.compile(optimizer=optimizers.Adam(),loss=losses.categorical_crossentropy,# sparse_categorical_crossentropy: beacuse the labels are integer and aren't one-hot  (help:https://stackoverflow.com/questions/54232549/multi-class-and-variable-size-image-classification-with-flow-from-directory) 
              metrics=['acc']) 

# model.build((None,197,275,3))

# Print Model
print(model.summary())

return model

而且，我的代码是：

#%% Call Pretrain VGG16 with ImageNet
conv_base = VGG16(weights='imagenet',include_top=False,input_shape=(197,3))

#%% Create Model
model = Create_Model(conv_base) 

#%% Freeze Conv_base(VGG-19) For Train Top layers
# Freezing just top layers
conv_base.trainable = False

# Generate Data
train_datagen = ImageDataGenerator(
                rescale=1./255,rotation_range=40,width_shift_range=0.2,height_shift_range=0.2,shear_range=0.2,zoom_range=0.2,horizontal_flip=True,featurewise_center=True,# set input mean to 0 over the dataset
                samplewise_center=True,# set each sample mean to 0
                featurewise_std_normalization=True,# divide inputs by std of the dataset
                samplewise_std_normalization=True,# divide each input by its std
                fill_mode='nearest'
                )

# preprocessing_function = preprocess,test_datagen =  ImageDataGenerator(
                rescale=1./255,# divide each input by its std
                )


train_generator = train_datagen.flow_from_directory(train_dir,target_size=(197,275),batch_size=16,seed=124000,# If the out of memory,reduce this parameter and increse "steps_per_epoch" while manitaining ratio
                                                    class_mode='categorical')                 # sparse: beacuse labels is integer and is not one-hot   (help:https://stackoverflow.com/questions/54232549/multi-class-and-variable-size-image-classification-with-flow-from-directory) 

validation_generator = test_datagen.flow_from_directory(validation_dir,reduce this parameter and increse "validation_step" while manitaining ratio
                                                        class_mode='categorical')             # sparse: beacuse labels is integer and is not one-hot   (help:https://stackoverflow.com/questions/54232549/multi-class-and-variable-size-image-classification-with-flow-from-directory)   


#%% Checkpoint
file_path = "weights.best.hdf5"
checkpoint = ModelCheckpoint(file_path,monitor='val_acc',verbose=1,save_best_only=True,mode='max')
callbacks_list = [checkpoint]

#%% Fit Model
'''
# Train Data:           176303
# Test  Data:           48974
# Validation Data:      19590
'''
# history = model.fit_generator(train_generator,steps_per_epoch=5510,# Warning: for batch_size = 32
#                               validation_data=validation_generator,#                               validation_steps=613,epochs=100,#                               callbacks=callbacks_list)

history = model.fit_generator(train_generator,steps_per_epoch=11020,# Warning: for batch_size = 16
                              validation_data=validation_generator,validation_steps=1251,callbacks=callbacks_list)


# history = model.fit_generator(train_generator,steps_per_epoch=2,# Warning: for Test
#                               validation_data=validation_generator,#                               validation_steps=1,epochs=5,#                               callbacks=callbacks_list)

#%% Test model
test_generator = test_datagen.flow_from_directory(test_dir,batch_size=20,class_mode='categorical') # sparse: beacuse labels is integer and is not one-hot   (help:https://stackoverflow.com/questions/54232549/multi-class-and-variable-size-image-classification-with-flow-from-directory) 

请帮帮我。

ValueError：检查目标时出错：预期dense_4具有形状（无，1082）但得到形状为（16，1802）的数组

解决方法

暂无找到可以解决该程序问题的有效方法，小编努力寻找整理中！

如果你已经找到好的解决方法，欢迎将解决方案带上本链接一起发送给小编。

小编邮箱:dio#foxmail.com (将#修改为@）