检查输入时出错：预期 dense_28_input 有 2 维，但得到的数组形状为 (3084, 32, 32)

Question

我正在尝试使用我的模型进行预测，其中数组的形状为 (3084, 32, 32)。 获取值错误 here is error image

这是我的模型

model.add(Dense(1028, input_shape = (3084,), activation = "sigmoid"))
model.add(Dense(514, activation="sigmoid"))
model.add(Dense(len(lb.classes_), activation="softmax"))

总结

Model: "sequential_21"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense_57 (Dense)             (None, 1028)              3171380   
_________________________________________________________________
dense_58 (Dense)             (None, 514)               528906    
_________________________________________________________________
dense_59 (Dense)             (None, 4)                 2060      
=================================================================
Total params: 3,702,346
Trainable params: 3,702,346
Non-trainable params: 0
_________________________________________________________________

尝试使用

opt = SGD(lr = 0.01)
model.compile(loss = "categorical_crossentropy", optimizer=opt, metrics=["accuracy"])
H = model.fit(train_X, train_Y, validation_data = (test_X, test_Y), epochs = 75, batch_size = 32)

Answer 1

您需要正确指定输入形状，下面的模型应该可以工作。

from tensorflow.keras.layers import *
from tensorflow.keras.models import *

model = Sequential()
model.add(Dense(1028, input_shape = (32,32), activation = "sigmoid"))
model.add(Flatten())
model.add(Dense(514, activation="sigmoid"))
model.add(Dense(4, activation="softmax"))

model.summary()