1 通道图像分类的 UnimplementedError

Question

我是 CNN 新手，遇到错误 UnimplementedError: Fused conv implementation does not support grouped convolutions for now. [[node sequential_3/activation_15/Relu (defined at <ipython-input-37-0f4d43f688ae>:5) ]] [Op:__inference_test_function_5422]

我正在遵循 this video 的代码，为 3 channel images 工作，但我需要使用 1 channel 图像，但它不起作用。以下是我的代码，我将特定行更改为-

model.add(Conv2D(32, (3, 3), input_shape = (256, 256, 1))) 还添加了 color_mode = 'grayscale' 在 train_generator = train_datagen.flow_from_directory(train_dir, target_size=(img_width, img_height), color_mode = 'grayscale', batch_size=batch_size, class_mode='binary')

但是，没有摆脱错误。有人可以帮我弄清楚吗？完整代码如下-

##import libraries here

img_width, img_height = 256, 256
train_dir = './train'
valid_dir = './validation'
train_samples = 35
valid_samples = 15
epochs = 7
batch_size = 5

if K.image_data_format() == 'channels_first':
  shape = (1, img_width, img_height)
else:
  shape = (img_width, img_height, 1)

train_datagen = ImageDataGenerator(rescale=1. / 255, shear_range=0.2, zoom_range=0.2, horizontal_flip=True)
test_datagen = ImageDataGenerator(rescale=1. / 255)
train_generator = train_datagen.flow_from_directory(train_dir, target_size=(img_width, img_height), batch_size=batch_size, class_mode='binary')
valid_generator = test_datagen.flow_from_directory(valid_dir, target_size=(img_width, img_height), 
                                                   batch_size=batch_size, class_mode='binary')

model = Sequential()
model.add(Conv2D(32, (3, 3), input_shape = shape))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2,2)))
model.add(Conv2D(32, (3, 3)))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2,2)))
## add more layers if you want
model.add(Flatten())
model.add(Dense(64))
model.add(Activation('relu'))
model.add(Dropout(0.5))
model.add(Dense(1))
model.add(Activation('sigmoid'))
model.summary()

model.compile(loss='binary_crossentropy', optimizer='rmsprop', metrics=['accuracy'])

model.fit_generator(train_generator, 
                    steps_per_epoch=train_samples // batch_size, 
                    epochs=epochs, 
                    validation_data=valid_generator, 
                    validation_steps=valid_samples // batch_size)

Answer 1

问题可能是由于模型中定义的通道数与数据集中的实际通道数不一致造成的。我建议在将图像输入模型之前将它们显式转换为灰度：

import tensorflow as tf
import pathlib
import matplotlib.pyplot as plt

dataset_url = "https://storage.googleapis.com/download.tensorflow.org/example_images/flower_photos.tgz"
data_dir = tf.keras.utils.get_file('flower_photos', origin=dataset_url, untar=True)
data_dir = pathlib.Path(data_dir)

img_gen = tf.keras.preprocessing.image.ImageDataGenerator(rescale=1. / 255, shear_range=0.2, zoom_range=0.2, horizontal_flip=True)

train_ds = tf.data.Dataset.from_generator(
    lambda: img_gen.flow_from_directory(data_dir, batch_size=32, shuffle=True),
    output_types=(tf.float32, tf.float32))

def convert_to_grayscale(image, label):
  return tf.image.rgb_to_grayscale(image), label

images, _ = next(iter(train_ds.take(1)))
image = images[0]
print('Before conversion --> ', image.shape)

train_ds = train_ds.map(convert_to_grayscale)
images, _ = next(iter(train_ds.take(1)))

image = images[0]
print('After conversion --> ', image.shape)

Found 3670 files belonging to 5 classes.
Using 2936 files for training.
Before conversion -->  (256, 256, 3)
After conversion -->  (256, 256, 1)

这是一个完整的工作示例：

import tensorflow as tf
import pathlib
import matplotlib.pyplot as plt

dataset_url = "https://storage.googleapis.com/download.tensorflow.org/example_images/flower_photos.tgz"
data_dir = tf.keras.utils.get_file('flower_photos', origin=dataset_url, untar=True)
data_dir = pathlib.Path(data_dir)

img_gen = tf.keras.preprocessing.image.ImageDataGenerator(rescale=1. / 255)
train_generator = img_gen.flow_from_directory(data_dir, target_size=(256, 256),
    batch_size=32,
    class_mode='sparse')

def convert_to_grayscale(images, labels):
  return tf.image.rgb_to_grayscale(images), labels

train_ds = tf.data.Dataset.from_generator(lambda: train_generator, output_types=(tf.float32, tf.float32))

train_ds = train_ds.map(convert_to_grayscale)

num_classes = 5

model = tf.keras.Sequential([
  tf.keras.layers.Conv2D(16, 3, padding='same', activation='relu', input_shape=(256, 256, 1)),
  tf.keras.layers.MaxPooling2D(),
  tf.keras.layers.Conv2D(32, 3, padding='same', activation='relu'),
  tf.keras.layers.MaxPooling2D(),
  tf.keras.layers.Conv2D(64, 3, padding='same', activation='relu'),
  tf.keras.layers.MaxPooling2D(),
  tf.keras.layers.Flatten(),
  tf.keras.layers.Dense(128, activation='relu'),
  tf.keras.layers.Dense(num_classes)
])

model.compile(optimizer='adam',
              loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True))

model.fit(train_ds, epochs=10)

1 通道图像分类的 UnimplementedError

UnimplementedError with 1 channel image classification

python

tensorflow

keras

deep-learning

image-classification