def inputs(batch_size,
eval_data='test_batch',
data_dir='../../cifardataset/cifar-10-batches-bin'):
"""Construct input for CIFAR evaluation using the Reader ops.
Args:
eval_data: bool, indicating if one should use the train or eval data set.
Returns:
images: Images. 4D tensor of [batch_size, IMAGE_SIZE, IMAGE_SIZE, 3] size.
labels: Labels. 1D tensor of [batch_size] size.
Raises:
ValueError: If no data_dir
"""
images, labels = cifar10_input.inputs(eval_data=eval_data,
data_dir=data_dir,
batch_size=batch_size)
images = tf.image.resize_images(
tf.cast(images, tf.float32),
tf.convert_to_tensor([64, 64], dtype=tf.int32))
labels = tf.one_hot(tf.cast(labels, tf.int32), depth=10, dtype=tf.int32)
return (images, labels)
def inputs(eval_data):
if not DATA_DIR:
raise ValueError('Please supply a data_dir')
data_dir = os.path.join(DATA_DIR, 'cifar-10-batches-bin')
return cifar10_input.inputs(eval_data=eval_data,
data_dir=data_dir,
batch_size=BATCH_SIZE)
def inputs(eval_data):
"""Construct input for CIFAR evaluation using the Reader ops.
Args:
eval_data: bool, indicating if one should use the train or eval data set.
Returns:
images: Images. 4D tensor of [batch_size, IMAGE_SIZE, IMAGE_SIZE, 3] size.
labels: Labels. 1D tensor of [batch_size] size.
Raises:
ValueError: If no data_dir
"""
if not FLAGS.data_dir:
raise ValueError('Please supply a data_dir')
data_dir = os.path.join(FLAGS.data_dir, 'cifar-10-batches-bin')
images, labels = cifar10_input.inputs(eval_data=eval_data,
data_dir=data_dir,
batch_size=FLAGS.batch_size)
if FLAGS.use_fp16:
images = tf.cast(images, tf.float16)
labels = tf.cast(labels, tf.float16)
else:
images = tf.cast(images, tf.float32)
labels = tf.cast(labels, tf.float32)
return images, labels
def inputs(eval_data):
"""Construct input for CIFAR evaluation using the Reader ops.
Args:
eval_data: bool, indicating if one should use the train or eval data set.
Returns:
images: Images. 4D tensor of [batch_size, IMAGE_SIZE, IMAGE_SIZE, 3] size.
labels: Labels. 1D tensor of [batch_size] size.
Raises:
ValueError: If no data_dir
"""
if not FLAGS.data_dir:
raise ValueError('Please supply a data_dir')
data_dir = os.path.join(FLAGS.data_dir, 'cifar-10-batches-bin')
return cifar10_input.inputs(eval_data=eval_data, data_dir=data_dir,
batch_size=FLAGS.batch_size)
def inputs(eval_data):
if not FLAGS.data_dir:
raise ValueError('Please supply a data_dir')
data_dir = os.path.join(FLAGS.data_dir, 'cifar-10-batches-bin')
return cifar10_input.inputs(eval_data=eval_data, data_dir=data_dir,
batch_size=FLAGS.batch_size)
# 初始化weight函数（附加L2的loss（L2正则化））
# L1正则：制造稀疏特征，大部分无用特征权重置0
# L2正则：特征权重不过大，特征权重比较平均
# 奥卡姆剃刀法则
cifar10.maybe_download_and_extract()
# 下载数据并解压展开
with tf.name_scope('Get'):
images_train, labels_train = cifar10_input.distorted_inputs(data_dir=data_dir, batch_size=batch_size)
'''
distorted_inputs
产生训练需要使用的数据（特征，label）
进行数据增强操作（图片随机水平旋转、随机剪切、随机设置亮度和对比度、数据标准化）
'''
images_test, labels_test = cifar10_input.inputs(eval_data=True, data_dir=data_dir, batch_size=batch_size)
# inputs：生成测试数据（只需进行24X24裁剪+数据标准化）
with tf.name_scope('Inputs'):
image_holder = tf.placeholder(tf.float32, [batch_size, 24, 24, 3], name='image_input')
label_holder = tf.placeholder(tf.int32, [batch_size], name='label_input')
'''
创建输入数据的placeholder（特征+label）
（batch_size后面定义网格结构时用到->样本条数需预先设定）
剪裁后图片大小24X24
颜色通道3（RGB）
'''
with tf.name_scope('Conv1'):
weight1 = variable_with_weight_loss(shape=[5, 5, 3, 64], stddev=5e-2, wl=0.0, name='W1')
with tf.name_scope('kernel1'):
kernel1 = tf.nn.conv2d(image_holder, weight1, [1, 1, 1, 1], padding='SAME', name='kernel1')
bias1 = tf.Variable(tf.constant(0.0, shape=[64]), name='b1')
def inputs(eval_data):
if not DATA_DIR:
raise ValueError('Please supply a data_dir')
data_dir = os.path.join(DATA_DIR, 'cifar-10-batches-bin')
return cifar10_input.inputs(eval_data=eval_data, data_dir=data_dir,
batch_size=BATCH_SIZE)
