tf识别单张图片ocr（0到9的识别）

pip install numpy -i http://pypi.douban.com/simple/ --trusted-host pypi.douban.com
pip install tensorflow-gpu==1.15.0 -i http://pypi.douban.com/simple/ --trusted-host pypi.douban.com
pip install opencv-python -i http://pypi.douban.com/simple/ --trusted-host pypi.douban.com

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

import time
 
import tensorflow as tf
import cv2 as cv
import numpy as np
 
 
def generate_image(a, rnd_size=100):
    image = np.zeros([28, 28], dtype=np.uint8)
    cv.putText(image, str(a), (7, 21), cv.FONT_HERSHEY_PLAIN, 1.3, 255, 2, 8)
 
    for i in range(rnd_size):
        row = np.random.randint(0, 28)
        col = np.random.randint(0, 28)
        image[row, col] = 0
 
    data = np.reshape(image, [1, 784])
    return image, data / 255
 
 
def display_images(images):
    import matplotlib.pyplot as plt
    size = len(images)
    for i in range(size):
        plt.subplot(2, 5, i + 1)
        plt.imshow(images[i])
 
    plt.show()
 
 
def load_data(sess, rnd_size=100, should_display_images=False):
    zero_image, zero = generate_image(0, rnd_size)
    one_image, one = generate_image(1, rnd_size)
    two_image, two = generate_image(2, rnd_size)
    three_image, three = generate_image(3, rnd_size)
    four_image, four = generate_image(4, rnd_size)
    five_image, five = generate_image(5, rnd_size)
    six_image, six = generate_image(6, rnd_size)
    seven_image, seven = generate_image(7, rnd_size)
    eight_image, eight = generate_image(8, rnd_size)
    nine_image, nine = generate_image(9, rnd_size)
 
    if should_display_images is True:
        display_images(
            [zero_image, one_image, two_image, three_image, four_image, five_image, six_image, seven_image, eight_image,
             nine_image])
 
    x_features = [zero, one, two, three, four, five, six, seven, eight, nine]
    x_features = np.array(x_features)
    x_features = np.reshape(x_features, (-1, 784))
 
    y = None
    y_lables = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
    y = sess.run(tf.one_hot(y_lables, 10))
 
    return x_features, y
 
 
def build_network(nhidden, classes_count):
    x = tf.placeholder(tf.float32, shape=[None, 784], name='x')
    y = tf.placeholder(tf.float32, shape=[None, classes_count], name='y')
 
    W1 = tf.Variable(tf.random_normal([784, nhidden]))
    b1 = tf.Variable(tf.random_normal([1, nhidden]))
    hidden1 = tf.add(tf.matmul(x, W1), b1)
    hidden1_result = tf.sigmoid(hidden1)
 
    W2 = tf.Variable(tf.random_normal([nhidden, classes_count]))
    b2 = tf.Variable(tf.random_normal([1, classes_count]))
    out = tf.add(tf.matmul(hidden1_result, W2), b2)
    out_result = tf.sigmoid(out)
 
    diff = tf.subtract(out_result, y)
    loss = tf.reduce_sum(tf.square(diff))
    train = tf.train.GradientDescentOptimizer(0.1)
    step = train.minimize(loss)
 
    tf.summary.scalar("loss", loss)
 
    return x, y, out_result, loss, step
 
 
def do_train():
    x, y, y_, loss, step = build_network(10, 10)
 
    with tf.Session() as sess:
        sess.run(tf.global_variables_initializer())
        summary_merged = tf.summary.merge_all()
        writer = tf.summary.FileWriter('logs-'+str(time.time()), sess.graph)
        for i in range(800):
            x_features, y_labels = load_data(sess)
            sess.run(step, feed_dict={x: x_features, y: y_labels})
            if (i + 1) % 50 == 0:
                cur_loss, summary_ = sess.run([loss, summary_merged], feed_dict={x: x_features, y: y_labels})
                writer.add_summary(summary_, i)
 
                pred_ys = sess.run(y_, feed_dict={x: x_features, y: y_labels})
                ys = tf.argmax(pred_ys, 0)
                ys_correct = tf.argmax(y_labels, 0)
 
                c = tf.equal(ys, ys_correct)
                count = tf.reduce_sum(tf.cast(c, tf.float32))
 
                r = sess.run(count)
                print(i + 1, ': loss: ', cur_loss, '正确个数：', r)
 
        print('*************************')
        x_features, y_labels = load_data(sess, 150, should_display_images=True)
        pred_ys = sess.run(y_, feed_dict={x: x_features})
        ys = tf.argmax(pred_ys, 0)
        r = sess.run(ys)
        print('图片识别结果：', r)
        writer.close()
 
 
if __name__ == '__main__':
    do_train()

50 : loss:  7.3588676 正确个数： 4.0
100 : loss:  6.6502814 正确个数： 5.0
150 : loss:  5.26784 正确个数： 7.0
200 : loss:  4.0591483 正确个数： 9.0
250 : loss:  3.4379258 正确个数： 8.0
300 : loss:  3.114149 正确个数： 8.0
350 : loss:  2.0274947 正确个数： 9.0
400 : loss:  1.4823446 正确个数： 10.0
450 : loss:  1.4051719 正确个数： 10.0
500 : loss:  0.91150457 正确个数： 10.0
550 : loss:  0.7835213 正确个数： 10.0
600 : loss:  0.72512466 正确个数： 10.0
650 : loss:  0.56525075 正确个数： 10.0
700 : loss:  0.4699742 正确个数： 10.0
750 : loss:  0.45453963 正确个数： 10.0
800 : loss:  0.45089394 正确个数： 10.0
*************************
图片识别结果： [0 1 2 3 4 5 6 7 8 9]