keras imagedatagenerator flow_from_dataframe：找到0个已验证的图像文件名属于0类

Question

我正在处理存储为.npy文件的3D图像数据（128,128,128）。当我尝试使用keras imagedatagenerator时，它会引发以下错误：

找到0个已验证的属于0 class的图像文件名。

我的代码如下：

processed_data =  r'C:/Users/dush/Desktop/Project/preprocessed'


from tensorflow.keras.preprocessing.image import ImageDataGenerator

img_size = 64
batch_size = 32

data_gen = ImageDataGenerator(horizontal_flip = True,
                              validation_split=0.2,
                               fill_mode = "nearest",
                               zoom_range = 0.3,
                               width_shift_range = 0.1,
                               height_shift_range = 0.1,
                               rotation_range = 30)


train_gen = data_gen.flow_from_dataframe(
        dataframe = train,
        directory = processed_data, 
        x_col = 'id',
        y_col = 'category',  
        target_size=(img_size, img_size, img_size),
        batch_size = batch_size,
        class_mode="binary",
        #validate_filenames=False,
        subset='training', seed = 23) #image_generator.flow_from_dataframe

valid_gen = data_gen.flow_from_dataframe(
    dataframe = train,
    directory = processed_data, 
    x_col = 'id',
    y_col = 'category',
    target_size=(img_size, img_size, img_size),
    #validate_filenames=False,
    batch_size=batch_size,
    class_mode="binary",
    subset='validation', shuffle=False, seed=23) 

如果我设置validate_filenames = false，则标识文件名但在拟合模型时会出现错误。我在做什么错？