CNN+LSTM序列X光图像分类实现报错：输入形状不兼容

尝试用CNN+LSTM网络处理预处理为150×150×3的X光图像序列，完成4分类任务。X_train形状为(4067, 150, 150, 3)，执行model.fit()时触发输入形状不兼容的错误。

代码实现： #

# x_train = np.reshape(x_train, (4067, 150, 150, 3))
# y_train = np.reshape(y_train, (4067, 4))

model = Sequential()
model.add(TimeDistributed(Conv2D(filters = 32, 
                                 kernel_size=(3,3), 
                                 padding='same', 
                                 activation = 'relu'), 
                          input_shape=(None, 150, 150, 3)))
model.add(TimeDistributed(AveragePooling2D()))
model.add(TimeDistributed(Flatten()))
model.add(LSTM(100))
model.add(Dense(24, activation='relu',name='output'))
model.add(Dense(4, activation = 'softmax'))

from tensorflow.keras.optimizers import Adam

optimizer = Adam(lr=0.001)
model.compile(optimizer = optimizer, 
                       loss = 'categorical_crossentropy',
                       metrics=['accuracy'])
from tensorflow.keras.callbacks import ReduceLROnPlateau

reduce_lr = ReduceLROnPlateau(monitor = 'val_accuracy',
                            factor = 0.3, 
                            patience = 2,
                            min_delta = 0.001,
                            mode = 'auto',
                            verbose = 1)

hist_cnn_lstm = model.fit(x_train, y_train, batch_size=64, epochs=15,
                                   validation_data = (x_valid, y_valid),
                          callbacks=reduce_lr
                         )

错误信息： #

Epoch 1/15
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-24-3ec61fbabcf1> in <module>()
      1 hist_cnn_lstm = model.fit(x_train, y_train, batch_size=64, epochs=15,
      2                                    validation_data = (x_valid, y_valid),
----> 3                           callbacks=reduce_lr
      4                          )

1 frames
/usr/local/lib/python3.7/dist-packages/tensorflow/python/framework/func_graph.py in autograph_handler(*args, **kwargs)
   1145           except Exception as e:  # pylint:disable=broad-except
   1146             if hasattr(e, "ag_error_metadata"):
-> 1147               raise e.ag_error_metadata.to_exception(e)
   1148             else:
   1149               raise

ValueError: in user code:

    File "/usr/local/lib/python3.7/dist-packages/keras/engine/training.py", line 1021, in train_function  *
        return step_function(self, iterator)
    File "/usr/local/lib/python3.7/dist-packages/keras/engine/training.py", line 1010, in step_function  **
        outputs = model.distribute_strategy.run(run_step, args=(data,))
    File "/usr/local/lib/python3.7/dist-packages/keras/engine/training.py", line 1000, in run_step  **
        outputs = model.train_step(data)
    File "/usr/local/lib/python3.7/dist-packages/keras/engine/training.py", line 859, in train_step
        y_pred = self(x, training=True)
    File "/usr/local/lib/python3.7/dist-packages/keras/utils/traceback_utils.py", line 67, in error_handler
        raise e.with_traceback(filtered_tb) from None
    File "/usr/local/lib/python3.7/dist-packages/keras/engine/input_spec.py", line 264, in assert_input_compatibility
        raise ValueError(f'Input {input_index} of layer "{layer_name}" is 

    ValueError: Input 0 of layer "sequential_1" is incompatible with the layer: expected shape=(None, None, 150, 150, 3), found shape=(None, 150, 150, 3)

解决方法 #

问题根源 #

TimeDistributed层用于处理序列数据，要求输入必须包含时间步维度。你的X_train形状是(样本数, 150, 150, 3)，缺少时间步维度；而模型输入定义的input_shape=(None, 150, 150, 3)对应完整输入形状是(样本数, 时间步数, 150, 150, 3)，两者不匹配导致报错。

具体修改步骤 #

1. 调整输入数据形状：给X_train和X_valid增加时间步维度。如果每个样本是单张图像（时间步数为1），用np.expand_dims扩展维度：

   x_train = np.expand_dims(x_train, axis=1)  # 形状变为(4067, 1, 150, 150, 3)
   x_valid = np.expand_dims(x_valid, axis=1)
   

   如果每个样本是包含N张图像的序列，需确保原始数据时间步维度正确，调整为(样本数, N, 150, 150, 3)。

2. （可选）明确输入形状的时间步：如果序列长度固定（比如每个样本都是1张图像），可把input_shape中的None改成固定值，让模型定义更清晰：

   model.add(TimeDistributed(Conv2D(filters=32, 
                                    kernel_size=(3,3), 
                                    padding='same', 
                                    activation='relu'), 
                             input_shape=(1, 150, 150, 3)))
   

修改后的完整代码示例 #

import numpy as np
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import TimeDistributed, Conv2D, AveragePooling2D, Flatten, LSTM, Dense
from tensorflow.keras.optimizers import Adam
from tensorflow.keras.callbacks import ReduceLROnPlateau

# 调整输入数据形状
x_train = np.expand_dims(x_train, axis=1)
x_valid = np.expand_dims(x_valid, axis=1)

model = Sequential()
# 这里可以用固定时间步1，或者保留None支持可变长度序列
model.add(TimeDistributed(Conv2D(filters = 32, 
                                 kernel_size=(3,3), 
                                 padding='same', 
                                 activation = 'relu'), 
                          input_shape=(1, 150, 150, 3)))
model.add(TimeDistributed(AveragePooling2D()))
model.add(TimeDistributed(Flatten()))
model.add(LSTM(100))
model.add(Dense(24, activation='relu',name='output'))
model.add(Dense(4, activation = 'softmax'))

optimizer = Adam(lr=0.001)
model.compile(optimizer = optimizer, 
              loss = 'categorical_crossentropy',
              metrics=['accuracy'])

reduce_lr = ReduceLROnPlateau(monitor = 'val_accuracy',
                              factor = 0.3, 
                              patience = 2,
                              min_delta = 0.001,
                              mode = 'auto',
                              verbose = 1)

hist_cnn_lstm = model.fit(x_train, y_train, batch_size=64, epochs=15,
                          validation_data = (x_valid, y_valid),
                          callbacks=reduce_lr)

内容的提问来源于stack exchange，提问作者Blackcat