Question

我正在使用具有 10 个类的 Keras 构建以下多类模型。

train_dir      = '/Users/...'
validation_dir = '/Users/...'
test_dir       = '/Users/...'

train_image_generator = ImageDataGenerator(rescale=1./255)
train_data_gen = train_image_generator.flow_from_directory(directory=train_dir, shuffle=False, target_size=IMAGE_SHAPE, class_mode='categorical')

validation_image_generator = ImageDataGenerator(rescale=1./255)
validation_data_gen = validation_image_generator.flow_from_directory(directory=validation_dir, shuffle=False, target_size=IMAGE_SHAPE, class_mode='categorical')

test_image_generator = ImageDataGenerator(rescale=1./255)
test_data_gen = test_image_generator.flow_from_directory(directory=test_dir, shuffle=False, target_size=IMAGE_SHAPE, class_mode='categorical')

型号：

num_classes = len(train_data_gen.class_indices)
model = tf.keras.Sequential([
    hub.KerasLayer("https://tfhub.dev/google/imagenet/mobilenet_v2_100_224/",
                   trainable=False, input_shape=IMAGE_SHAPE+(3,)),
    tf.keras.layers.Dense(num_classes, activation='softmax')
])

model.summary()

train_steps_per_epoch = np.ceil(train_data_gen.samples/train_data_gen.batch_size)
val_steps_per_epoch = np.ceil(validation_data_gen.samples/validation_data_gen.batch_size)

epochs = 10
history = model.fit(train_data_gen, epochs=epochs,
                    steps_per_epoch=train_steps_per_epoch,
                    validation_data=validation_data_gen,
                    validation_steps=val_steps_per_epoch)

我想得到混淆矩阵，但由于维度不匹配而无法工作，或者我得到一个只有 1 和 0 的巨大矩阵。

y_classes = predictions.argmax(axis=0) # Axis change only results in an error

cm = confusion_matrix(y_classes.astype(str), np.array(dirlist_val))
cm

array([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0],
       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0],
       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0],
       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0],
       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]])

一定是出了什么问题。 dirlist_val 是一个列表，其中包含 10 个类的所有名称。所以它必须是一个 10x10 的矩阵。这个矩阵怎么这么大！

我这样做只是为了测试目的，所以不要怀疑任何设置。试图找出我的思维错误。

来自 Keras 多类模型的混淆矩阵

0 个答案: