Tensorflow Keras flow_来自_目录,未读取所有类中的所有图像

Tensorflow Keras flow_来自_目录,未读取所有类中的所有图像,tensorflow,keras,directory,computer-vision,google-colaboratory,Tensorflow,Keras,Directory,Computer Vision,Google Colaboratory,我在Google Colab中使用Keras ImageDataGenerator。 我有3类网球图像,如下所示: FRAMES_RGB |- bhnd --> contains about 34000 image files |- forehand --> contains about 34000 image files |- idle --> contains about 9082 image files FRAMES_RGB_VALIDATION |- b

我在Google Colab中使用Keras ImageDataGenerator。 我有3类网球图像,如下所示:

FRAMES_RGB

|- bhnd --> contains about 34000 image files  
|- forehand --> contains about 34000 image files  
|- idle --> contains about 9082 image files

FRAMES_RGB_VALIDATION  
|- bhnd --> contains about 3500 image files  
|- forehand --> contains about 3500 image files  
|- idle --> contains about 1000 image files
但是,当我运行代码时,会收到以下消息:

Reading training and validation data...
Found 9082 images belonging to 3 classes.
Found 8324 images belonging to 3 classes.
因此,我知道它只从训练集中的空闲类中获取图像(在验证集中一切正常),但它仍然知道有3个类

这些是我的发电机:

 training_datagen = tf.keras.preprocessing.image.ImageDataGenerator(
        preprocessing_function=self._preprocessing_function,
        rotation_range=30,
        width_shift_range=0.2,
        height_shift_range=0.2,
        shear_range=0.2,
        zoom_range=0.2,
        horizontal_flip=True,  # Randomly flip half of the images horizontally
        fill_mode='nearest'  # Strategy used for filling in new pixels that appear after transforming images
    )

    validation_datagen = tf.keras.preprocessing.image.ImageDataGenerator(preprocessing_function=self._preprocessing_function)

    training_generator = training_datagen.flow_from_directory(
        training_dir,
        target_size=self._target_size,
        batch_size=training_batch_size,
        class_mode='categorical'
    )

    validation_generator = validation_datagen.flow_from_directory(
        validation_dir,
        target_size=self._target_size,
        batch_size=validation_batch_size,
        class_mode='categorical',
        shuffle=False
    )

为什么它只读取我训练集中的空闲类?

似乎连验证数据都是错误的。您能否共享完整的代码来复制您的问题?因此我们可以尝试帮助您。@t这是指向我正在执行所有操作的文件夹的链接,笔记本是“Copia de training_thetis_non seq”。