Keras CNN-LSTM：制作 y_train 时出错

Question

这是我第一次在这里提问（这意味着我真的需要帮助），抱歉我的英语不好。我想在 Keras 中制作一个用于视频分类的 cnn-lstm 层，但我在制作 y_train 时遇到了问题。之后我会描述我的问题。我有视频数据集（1 个视频有 10 帧），我将视频转换为图像。首先，我将数据集拆分为 xtrain、xtest、ytrain 和 ytest（20% 测试，80% 训练）然后我做到了。

X_train, X_test = img_data[:trainco], img_data[trainco:]
y_train, y_test = y[:trainco], y[trainco:]

X_train 形状：(2280, 64, 64, 1) -> 我有 2280 张图片，64x64 高 x 宽，1 个通道

y_train 形状：(2280, 26) -> 26 类

然后我必须在进入 cnn-lstm 过程之前重塑它们。 *注意：我对 x_test 和 y_test

做同样的事情

time_steps = 10 (because I have 10 frames per video)

X_train = X_train.reshape(int(X_train.shape[0] / time_steps), time_steps, X_train.shape[1], X_train.shape[2], X_train.shape[3])
y_train = y_train.reshape(int(y_train.shape[0] / time_steps), time_steps, y_train.shape[1])

X_train 形状：(228, 10, 64, 64, 1), y_train 形状：(228, 10, 26)

然后这是我的模型：

model = Sequential()
model.add(TimeDistributed(Conv2D(32, (3, 3), strides=(2, 2), activation='relu', padding='same'), input_shape=X_train.shape[1:]))
model.add(TimeDistributed(MaxPooling2D((2, 2), strides=(2, 2))))
model.add(TimeDistributed(Conv2D(32, (3, 3), padding='same', activation='relu')))
model.add(TimeDistributed(MaxPooling2D((2, 2), strides=(2, 2))))
model.add(TimeDistributed(Flatten()))
model.add(LSTM(256, return_sequences=False, input_shape=(64, 64)))
model.add(Dense(128))
model.add(Dense(64))
model.add(Dense(num_classes, activation='softmax'))
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=["accuracy"])
checkpoint = ModelCheckpoint(fname, monitor='acc', verbose=1, save_best_only=True, mode='max', save_weights_only=True)
hist = model.fit(X_train, y_train, batch_size=num_batch, nb_epoch=num_epoch, verbose=1, validation_data=(X_test, y_test), callbacks=[checkpoint])

但我收到一条错误消息

ValueError: Error when checking target: expected dense_3 to have 2 dimensions, but got array with shape (228, 10, 26)

就像它说的那样，应该有 2 个维度。我将代码更改为

y_train = y_train.reshape(int(y_train.shape[0] / time_steps), y_train.shape[1])

我又收到一个错误提示

ValueError: cannot reshape array of size 59280 into shape (228,26)

然后我再次将代码更改为

y_train = y_train.reshape(y_train.shape[0], y_train.shape[1])

我仍然遇到错误

ValueError: Input arrays should have the same number of samples as target arrays. Found 228 input samples and 2280 target samples.

我该怎么办？我知道问题所在，但我不知道如何解决。请帮助我。

Answer 1

我根据您的情况重新创建了一个稍微简化的版本以重现该问题。基本上，LSTM 层似乎只针对整个时间步长序列输出一个结果，从而将输出中的维度从 3 减少到 2。如果您运行下面是我的程序，我添加了 model.summary() ，它提供了架构的详细信息。

from keras import Sequential
from keras.layers import TimeDistributed, Dense, Conv2D, MaxPooling2D, Flatten, LSTM
import numpy as np

X_train = np.random.random((228, 10, 64, 64, 1))
y_train = np.random.randint(2, size=(228, 10, 26))
num_classes = 26

# Create the model
model = Sequential()
model.add(TimeDistributed(Conv2D(32, (3, 3), strides=(2, 2), activation='relu', padding='same'), input_shape=X_train.shape[1:]))
model.add(TimeDistributed(MaxPooling2D((2, 2), strides=(2, 2))))
model.add(TimeDistributed(Conv2D(32, (3, 3), padding='same', activation='relu')))
model.add(TimeDistributed(MaxPooling2D((2, 2), strides=(2, 2))))
model.add(TimeDistributed(Flatten(),name='Flatten'))
model.add(LSTM(256, return_sequences=False, input_shape=(64, 64)))
model.add(Dense(128))
model.add(Dense(64))
model.add(Dense(num_classes, activation='softmax', name='FinalDense'))
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=["accuracy"])

#
model.summary()
# hist = model.fit(X_train, y_train, epochs=1)

我相信您需要决定是要降低 y_train（目标）数据的维度以与模型保持一致，还是更改模型。希望对您有所帮助。

Keras CNN-LSTM：制作 y_train 时出错

Keras CNN-LSTM : Error while making y_train

keras

conv-neural-network

lstm

recurrent-neural-network

train-test-split