Python，张量流回归

Question

我想通过 tensorflow 解决回归问题。 我想预测猫/狗的图片。

x = tf.placeholder(tf.float32, shape=[512, 512])  #this is input nnet
y_ = tf.placeholder(tf.float32, shape=[1, 1])  #this is output nnet


for i in range(24):
    img = get_image_from_file("./MOJE/koty_nauka/kot" + str(i + 1) +
                              ".jpg")
    out = y_conv.eval(feed_dict={
        x: img, y_: [[1]], keep_prob: 1.0})
    print("----")
    print(out)

而且输出总是一样的：

----
[[ 1.]
[ 1.]
[ 1.]
[ 1.]
[ 1.]
[ 1.]
[ 1.]
[ 1.]]

我的网络只返回这个值。是否可以正确学习nnet？

如果有猫 nnet 应该返回 1
如果有狗 nnet 应该返回 0
有可能吗？

Answer 1

获得所有 x (img) 数据后，您需要定义成本和训练操作：

w = init_weights([784, 10]) # #of x vec and # of labels 
py_x = tf.matmul(X, w)

cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(py_x, Y)) # compute mean cross entropy (softmax is applied internally)
train_op = tf.train.GradientDescentOptimizer(0.05).minimize(cost) # construct optimizer
predict_op = tf.argmax(py_x, 1) # at predict time, evaluate the argmax of the logistic regression

然后，就可以使用您的 x 输入来训练模型了：

# Launch the graph in a session
with tf.Session() as sess:
    # you need to initialize all variables
    tf.initialize_all_variables().run()

    for i in range(100):
        sess.run(train_op, feed_dict={X: x, Y: y) # feed your x data and label

请参阅https://github.com/nlintz/TensorFlow-Tutorials/blob/master/2_logistic_regression.py 上的完整代码示例。

Answer 2

这是我的错，我有这个代码，我不想粘贴所有无用的代码来做那个

"""Import."""
import tensorflow as tf
import cv2

sess = tf.InteractiveSession()
x = tf.placeholder(tf.float32, shape=[512, 512])
y_ = tf.placeholder(tf.float32, shape=[1, 1])


def get_image_from_file(file_name):
    """Function get_image_from_file."""
    return cv2.resize(cv2.imread(file_name, 0), (512, 512),
                      interpolation=cv2.INTER_CUBIC)


def weight_variable(shape):
    """Foo."""
    initial = tf.truncated_normal(shape, stddev=0.1)
    return tf.Variable(initial)


def bias_variable(shape):
    """Foo."""
    initial = tf.constant(0.1, shape=shape)
    return tf.Variable(initial)


def conv2d(x, w):
    """Foo."""
    return tf.nn.conv2d(x, w, strides=[1, 1, 1, 1], padding='SAME')


def max_pool_2x2(x):
    """Max pool 2x2."""
    return tf.nn.max_pool(x, ksize=[1, 2, 2, 1],
                          strides=[1, 2, 2, 1], padding='SAME')


if __name__ == "__main__":
    # 1st layer
    w_conv1 = weight_variable([5, 5, 1, 16])
    b_conv1 = bias_variable([16])

    x_image = tf.reshape(x, [-1, 512, 512, 1])

    h_conv1 = tf.nn.relu(conv2d(x_image, w_conv1) + b_conv1)
    h_pool1 = max_pool_2x2(h_conv1)

    # 2nd layer
    w_conv2 = weight_variable([5, 5, 16, 32])
    b_conv2 = bias_variable([32])

    h_conv2 = tf.nn.relu(conv2d(h_pool1, w_conv2) + b_conv2)
    h_pool2 = max_pool_2x2(h_conv2)

    # connection layer
    w_fc1 = weight_variable([32 * 32 * 64, 128])
    b_fc1 = bias_variable([128])

    h_pool2_flat = tf.reshape(h_pool2, [-1, 32 * 32 * 64])
    h_fc1 = tf.nn.relu(tf.matmul(h_pool2_flat, w_fc1) + b_fc1)

    # zapobieganie przeuczeniu
    keep_prob = tf.placeholder(tf.float32)
    h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob)

    # output layer
    w_fc2 = weight_variable([128, 1])
    b_fc2 = bias_variable([1])

    y_conv = tf.nn.softmax(tf.matmul(h_fc1_drop, w_fc2) + b_fc2)
    cross_entropy = -tf.reduce_sum(y_ * tf.log(y_conv))
    train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
    correct_prediction = tf.equal(tf.argmax(y_conv, 1), tf.argmax(y_, 1))
    accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
    sess.run(tf.initialize_all_variables())
    for i in range(24):
        img = get_image_from_file("./MOJE/koty_nauka/kot" + str(i + 1) +
                                  ".jpg")

        out = y_conv.eval(feed_dict={
            x: img, y_: [[1]], keep_prob: 1.0})
        print("----")
        print(out)