基于Tensorflow，以MINST为数据集训练手写数字识别模型，并调用打开摄像头的API，对拍摄图片中一个或多个数字进行识别

# -*- coding: utf-8 -*- """ Created on Sun Sep 9 22:55:06 2018 @author: Boyu """ from PIL import Image, ImageFilter import tensorflow as tf import matplotlib.pyplot as plt from number_cuts import * from image_process import * from capture_nums import * import cv2 tf.reset_default_graph() image_cap_and_process() a = eval(input('请输入数字的位数：')) #Picture captured by camera in function cap_numbers() path1 = 'cap_nums.png' #Picture after function :image_cap_and_process() path2 = 'binary.png' img1 = Image.open(path1) img2 = Image.open(path2) """print('原始图片如下：') plt.imshow(img0) plt.imshow(img1) plt.show()""" #print('二值化图片如下：') plt.imshow(img2) plt.show() cut_to_nums(img2,a,1) #def get_images(): # images = [] # for i in range(1,9): # locals()['im%i'% i] =[] # locals()['im%i'% i].append(Image.open('../single_nums/single_nums' + str(i-1)+'.png')) def imageprepare(): global a im1 = Image.open('../single_nums/single_nums0.png') im2 = Image.open('../single_nums/single_nums1.png') im3 = Image.open('../single_nums/single_nums2.png') im4 = Image.open('../single_nums/single_nums3.png') im5 = Image.open('../single_nums/single_nums4.png') im6 = Image.open('../single_nums/single_nums5.png') im7 = Image.open('../single_nums/single_nums6.png') im8 = Image.open('../single_nums/single_nums7.png') tv1 = list(im1.convert('L').getdata()) tv2 = list(im2.convert('L').getdata()) tv3 = list(im3.convert('L').getdata()) tv4 = list(im4.convert('L').getdata()) tv5 = list(im5.convert('L').getdata()) tv6 = list(im6.convert('L').getdata()) tv7 = list(im7.convert('L').getdata()) tv8 = list(im8.convert('L').getdata()) tva1 = [(255-x)*1.0/255.0 for x in tv1] tva2 = [(255-x)*1.0/255.0 for x in tv2] tva3 = [(255-x)*1.0/255.0 for x in tv3] tva4 = [(255-x)*1.0/255.0 for x in tv4] tva5 = [(255-x)*1.0/255.0 for x in tv5] tva6 = [(255-x)*1.0/255.0 for x in tv6] tva7 = [(255-x)*1.0/255.0 for x in tv7] tva8 = [(255-x)*1.0/255.0 for x in tv8] return tva1,tva2,tva3,tva4,tva5,tva6,tva7,tva8 """images = [im1,im2,im3,im4,im5,im6,im7,im8] for i in range(a): plt.subplot(1,a,i+1),plt.imshow(images[i]) plt.xticks([]),plt.yticks([]) plt.show()""" result=imageprepare() x = tf.placeholder(tf.float32, [None, 784]) y_ = tf.placeholder(tf.float32, [None, 10]) 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): return tf.nn.conv2d(x, W, strides = [1,1,1,1], padding = 'SAME') def max_pool_2x2(x): return tf.nn.max_pool(x, ksize=[1,2,2,1], strides=[1,2,2,1], padding='SAME') W_conv1 = weight_variable([5, 5, 1, 32]) b_conv1 = bias_variable([32]) x_image = tf.reshape(x,[-1,28,28,1]) #generate 28*28*32 h_conv1 = tf.nn.relu(conv2d(x_image,W_conv1) + b_conv1) #generate 14*14*32 h_pool1 = max_pool_2x2(h_conv1) W_conv2 = weight_variable([5, 5, 32, 64]) b_conv2 = bias_variable([64]) #generate 14*14*64 h_conv2 = tf.nn.relu(conv2d(h_pool1, W_conv2) + b_conv2) #generate 7*7*64 h_pool2 = max_pool_2x2(h_conv2) #input_nodes*output_nodes parameters W_fc1 = weight_variable([7 * 7 * 64, 1024]) #1024 bias b_fc1 = bias_variable([1024]) #flatten h_pool2_flat = tf.reshape(h_pool2, [-1, 7*7*64]) #work out full_connection2's all parameters h_fc1 = tf.nn.relu(tf.matmul(h_pool2_flat, W_fc1) + b_fc1) keep_prob = tf.placeholder("float") h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob) W_fc2 = weight_variable([1024, 10]) b_fc2 = bias_variable([10]) 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, "float")) saver = tf.train.Saver() with tf.Session() as sess: sess.run(tf.global_variables_initializer()) saver.restore(sess, "../MNIST_MODEL/model.ckpt") #使用模型，参数和之前的代码保持一致 prediction=tf.argmax(y_conv,1) predint1=prediction.eval(feed_dict={x: [result[0]],keep_prob: 1.0}, session=sess) predint2=prediction.eval(feed_dict={x: [result[1]],keep_prob: 1.0}, session=sess) predint3=prediction.eval(feed_dict={x: [result[2]],keep_prob: 1.0}, session=sess) predint4=prediction.eval(feed_dict={x: [result[3]],keep_prob: 1.0}, session=sess) predint5=prediction.eval(feed_dict={x: [result[4]],keep_prob: 1.0}, session=sess) predint6=prediction.eval(feed_dict={x: [result[5]],keep_prob: 1.0}, session=sess) predint7=prediction.eval(feed_dict={x: [result[6]],keep_prob: 1.0}, session=sess) predint8=prediction.eval(feed_dict={x: [result[7]],keep_prob: 1.0}, session=sess) print('识别结果:') if a == 1: print(predint1[0]) if a == 2: print(predint1[0],predint2[0]) if a == 3 : print(predint1[0],predint2[0],predint3[0]) if a == 4: print(predint1[0],predint2[0],predint3[0],predint4[0]) if a == 5: print(predint1[0],predint2[0],predint3[0],predint4[0],predint5[0]) if a == 6 : print(predint1[0],predint2[0],predint3[0],predint4[0],predint5[0],predint6[0]) if a == 7: print(predint1[0],predint2[0],predint3[0],predint4[0],predint5[0],predint6[0],predint7[0]) if a == 8 : print(predint1[0],predint2[0],predint3[0],predint4[0],predint5[0],predint6[0],predint7[0],predint8[0]) #../code_after.png