TensorFlow 总是返回相同的结果

问题描述

我正在尝试使用卷积神经网络进行分类，只有 2 个类，我没有看到我的输入图像或网络有任何问题，但我想知道为什么结果(准确度)总是返回相同的值?

我通过引用这个来构建我的模型:

谁能帮帮我??谢谢.

解决方案

我建议您规范化所有输入数据和标签.并确保训练数据和测试数据以相同的尺度进行归一化.

I am trying to do classification by using Convolution Neural Network, only 2 classes, I don't see my input images or the network has any problems, but I am wonder why the result (accuracy) always return me same value ?

I build my model by referring this :https://github.com/MorvanZhou/tutorials/blob/master/tensorflowTUT/tf18_CNN3/full_code.py

from __future__ import print_function
import tensorflow as tf
import matplotlib.image as mpimg
import matplotlib.pyplot as plt
def getTrainLabels():
    labels=[]
    file = open('data/Class1/Class1/Train/Label/Labels.txt', 'r')
    for line in file:
        if len(line)<=25:
            labels.append([0,1])
        else:
            labels.append([1,0])
    return labels

def getTrainImages():
    images = []
    for i in range(576,1151):#1151
        if i<1000:
            filename = 'data/Class1/Class1/Train/0'+str(i)+'.PNG'
            raw_image_data = mpimg.imread(filename)
            images.append(raw_image_data)
        else:
            filename = 'data/Class1/Class1/Train/'+str(i)+'.PNG'
            raw_image_data = mpimg.imread(filename)
            images.append(raw_image_data)
    # step 2
    return images



def getTestImages():
    images = []
    for i in range(1,576):
        if i<10:
            filename = 'data/Class1/Class1/Test/000'+str(i)+'.PNG'
            raw_image_data = mpimg.imread(filename)
            images.append(raw_image_data)
        elif i<100:
            filename = 'data/Class1/Class1/Test/00'+str(i)+'.PNG'
            raw_image_data = mpimg.imread(filename)
            images.append(raw_image_data)
        elif i<1000:
            filename = 'data/Class1/Class1/Test/0'+str(i)+'.PNG'
            raw_image_data = mpimg.imread(filename)
            images.append(raw_image_data)
        else:
            filename = 'data/Class1/Class1/Test/'+str(i)+'.PNG'
            raw_image_data = mpimg.imread(filename)
            images.append(raw_image_data)
    # step 2
    return images

def getTestLabels():
    labels=[]
    file = open('data/Class1/Class1/Test/Label/Labels.txt', 'r')
    for line in file:
        if len(line)<=25:
            labels.append([0,1])
        else:
            labels.append([1,0])
    return labels

def compute_accuracy(v_xs, v_ys):
    global prediction
    y_pre = sess.run(prediction, feed_dict={xs: v_xs, keep_prob: 1})
    correct_prediction = tf.equal(tf.argmax(y_pre,1), tf.argmax(v_ys,1))
    accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
    result = sess.run(accuracy, feed_dict={xs: v_xs, ys: v_ys, keep_prob: 1})
    return result

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

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

def conv2d(x, W):
    # stride [1, x_movement, y_movement, 1]
    # Must have strides[0] = strides[3] = 1
    return tf.nn.conv2d(x, W, strides=[1, 1, 1, 1], padding='SAME') #SAME or VALID

def max_pool_2x2(x):
    # stride [1, x_movement, y_movement, 1]
    return tf.nn.max_pool(x, ksize=[1,2,2,1], strides=[1,2,2,1], padding='SAME')

# define placeholder for inputs to network
xs = tf.placeholder(tf.float32, [None, 512, 512]) # 512x512
ys = tf.placeholder(tf.float32, [None,2])
keep_prob = tf.placeholder(tf.float32)
x_image = tf.reshape(xs, [-1, 512, 512, 1])
# print(x_image.shape)  # [n_samples, 512,512,1]

## conv1 layer ##
W_conv1 = weight_variable([5,5, 1,8]) # patch 5x5, in size 1, out size 32
b_conv1 = bias_variable([8])
h_conv1 = tf.nn.relu(conv2d(x_image, W_conv1) + b_conv1) # output size 512x512x32
h_pool1 = max_pool_2x2(h_conv1)                                         # output size 256x256x32

## conv2 layer ##
W_conv2 = weight_variable([5,5, 8, 8]) # patch 5x5, in size 32, out size 64
b_conv2 = bias_variable([8])
h_conv2 = tf.nn.relu(conv2d(h_pool1, W_conv2) + b_conv2) # output size 256x256x64
h_pool2 = max_pool_2x2(h_conv2)                                         # output size 128x128x64

## func1 layer ##
W_fc1 = weight_variable([128*128*8, 8])
b_fc1 = bias_variable([8])
# [n_samples, 7, 7, 64] ->> [n_samples, 7*7*64]
h_pool2_flat = tf.reshape(h_pool2, [-1, 128*128*8])
h_fc1 = tf.nn.relu(tf.matmul(h_pool2_flat, W_fc1) + b_fc1)
h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob)

## func2 layer ##
W_fc2 = weight_variable([8, 2]) # only 2 class, defect or defect-free
b_fc2 = bias_variable([2])
prediction = tf.nn.softmax(tf.matmul(h_fc1_drop, W_fc2) + b_fc2)


# the error between prediction and real data
cross_entropy = tf.reduce_mean(-tf.reduce_sum(ys * tf.log(prediction),
                                              reduction_indices=[1]))       # loss
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)

sess = tf.Session()
# important step
sess.run(tf.initialize_all_variables())

batch_xs = getTrainImages()
batch_ys = getTrainLabels()
test_images = getTestImages()
test_labels = getTestLabels()
m_oH = 0
m_oT = 5
for i in range(1,116):
    #batch_xs, batch_ys = mnist.train.next_batch(100)
    sess.run(train_step, feed_dict={xs: batch_xs[m_oH:m_oT], ys: batch_ys[m_oH:m_oT],keep_prob:1})
    m_oH=m_oH+5
    m_oT=m_oT+5
    if i % 50 == 0:
        print(compute_accuracy(
            test_images, test_labels))

print(compute_accuracy(test_images, test_labels))

Below is the result : It always return 0.876522

Can anyone help me ?? thanks.

解决方案

I reccomend you to normalize all input data and labels. And be sure that training data and test data are normalized in the same scale.

这篇关于TensorFlow 总是返回相同的结果的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持！