如何在 Keras 的 ResNet 152 V2 中使用 TFrecord

如何解决如何在 Keras 的 ResNet 152 V2 中使用 TFrecord

我打算将 8000 张图像分成一个 TFRecord 并通过将其分成训练、验证和测试来使用它。但是，我不确定如何在 Keras 原生支持的 ResNet152 V2 中使用 TFRecord。

我一直在 Input Shape 部分遇到错误，我尝试将 Shape 更改为 None 并将其置于原始图像大小 (256,256,3)。但总是错误被称为

这是我的代码

    import os
    import numpy as np
    from sklearn.model_selection import train_test_split,cross_val_score,StratifiedKFold
    from tensorflow.python.keras.callbacks import TensorBoard
    from keras.applications.resnet_v2 import ResNet152V2
    from keras import optimizers
    from keras.callbacks import ModelCheckpoint,History
    from keras.layers import BatchNormalization,Flatten,Dense,Dropout
    from keras.models import Model
    import tensorflow as tf
    from keras import metrics
    from IPython.core import display


    def Load_TFrecord(TFrecord):
        train_size = int(0.6 * 8000)
        val_size = int(0.2 * 8000)
        test_size = int(0.2 * 8000)

        full_dataset = tf.data.TFRecordDataset(TFrecord)
        train_dataset = full_dataset.take(train_size)
        test_dataset = full_dataset.skip(train_size)
        val_dataset = test_dataset.skip(val_size)
        test_dataset = test_dataset.take(test_size)
        return train_dataset,test_dataset,val_dataset




    def Train(train,val,test):

        API_model = ResNet152V2(include_top=False,weights=None,input_tensor=None,input_shape=None,pooling=None,classes=8)
        flat1 = Flatten()(API_model.layers[-1].output)
        dropout1 = Dropout(0.5)(flat1)
        class1 = Dense(1024,activation='relu')(dropout1)
        dropout2 = Dropout(0.5)(class1)
        output = Dense(8,activation='softmax')(dropout2)
        model = Model(inputs=API_model.inputs,outputs=output)
        sgd = optimizers.Adagrad(lr=0.01,epsilon=None,decay=1e-6)
        model.summary()
        print(len(model.layers))
        model.compile(loss="categorical_crossentropy",optimizer=sgd,metrics=[metrics.mae,metrics.categorical_accuracy])
        tensorboard = TensorBoard(log_dir="logs_test/test_1",histogram_freq=1,write_graph=True,write_images=True)
        model_checkpoint = ModelCheckpoint('test_.hdf5',monitor='loss',save_best_only=True)
        model.fit(train,validation_data=val,batch_size=40,epochs=100,callbacks=[tensorboard,model_checkpoint])


    if __name__ == '__main__':
        gpus = tf.config.experimental.list_physical_devices('GPU')
        if gpus:
            try:
                tf.config.experimental.set_memory_growth(gpus[0],True)
            except RuntimeError as e:
                print(e)
        train_set,test_set,val_set = Load_TFrecord("Endo_Dataset.tfrecords")
        Train(train_set,val_set)

这是我得到的错误：

    Traceback (most recent call last):
      File "test_TFrecord_Train.py",line 60,in <module>
    Train(train_set,val_set)
      File "test_TFrecord_Train.py",line 37,in Train
    class1 = Dense(1024,activation='relu')(dropout1)
      File "/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/engine/base_layer.py",line 952,in __call__
    input_list)
      File "/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/engine/base_layer.py",line 1091,in _functional_construction_call
    inputs,input_masks,args,kwargs)
      File "/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/engine/base_layer.py",line 822,in _keras_tensor_symbolic_call
    return self._infer_output_signature(inputs,kwargs,input_masks)
      File "/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/engine/base_layer.py",line 862,in _infer_output_signature
    self._maybe_build(inputs)
      File "/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/engine/base_layer.py",line 2710,in _maybe_build
    self.build(input_shapes)  # pylint:disable=not-callable
      File "/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/layers/core.py",line 1182,in build
    raise ValueError('The last dimension of the inputs to `Dense` '
    ValueError: The last dimension of the inputs to `Dense` should be defined. Found `None`.