您好,登錄后才能下訂單哦!
這篇文章主要為大家展示了“TensorFlow如何實(shí)現(xiàn)車牌識(shí)別功能”,內(nèi)容簡(jiǎn)而易懂,條理清晰,希望能夠幫助大家解決疑惑,下面讓小編帶領(lǐng)大家一起研究并學(xué)習(xí)一下“TensorFlow如何實(shí)現(xiàn)車牌識(shí)別功能”這篇文章吧。
如何使用TensorFlow進(jìn)行車牌識(shí)別,但是,當(dāng)時(shí)采用的數(shù)據(jù)集是MNIST數(shù)字手寫體,只能分類0-9共10個(gè)數(shù)字,無(wú)法分類省份簡(jiǎn)稱和字母,局限性較大,無(wú)實(shí)際意義。
經(jīng)過(guò)圖像定位分割處理,博主收集了相關(guān)省份簡(jiǎn)稱和26個(gè)字母的圖片數(shù)據(jù)集,結(jié)合前述博文中貼出的python+TensorFlow代碼,實(shí)現(xiàn)了完整的車牌識(shí)別功能。本著分享精神,在此送上全部代碼和車牌數(shù)據(jù)集。
車牌數(shù)據(jù)集下載地址(約4000張圖片):tf_car_license_dataset_jb51.rar
省份簡(jiǎn)稱訓(xùn)練+識(shí)別代碼(保存文件名為train-license-province.py)(拷貝代碼請(qǐng)務(wù)必注意python文本縮進(jìn),只要有一處縮進(jìn)錯(cuò)誤,就無(wú)法得到正確結(jié)果,或者出現(xiàn)異常):
#!/usr/bin/python3.5 # -*- coding: utf-8 -*- import sys import os import time import random import numpy as np import tensorflow as tf from PIL import Image SIZE = 1280 WIDTH = 32 HEIGHT = 40 NUM_CLASSES = 6 iterations = 300 SAVER_DIR = "train-saver/province/" PROVINCES = ("京","閩","粵","蘇","滬","浙") nProvinceIndex = 0 time_begin = time.time() # 定義輸入節(jié)點(diǎn),對(duì)應(yīng)于圖片像素值矩陣集合和圖片標(biāo)簽(即所代表的數(shù)字) x = tf.placeholder(tf.float32, shape=[None, SIZE]) y_ = tf.placeholder(tf.float32, shape=[None, NUM_CLASSES]) x_image = tf.reshape(x, [-1, WIDTH, HEIGHT, 1]) # 定義卷積函數(shù) def conv_layer(inputs, W, b, conv_strides, kernel_size, pool_strides, padding): L1_conv = tf.nn.conv2d(inputs, W, strides=conv_strides, padding=padding) L1_relu = tf.nn.relu(L1_conv + b) return tf.nn.max_pool(L1_relu, ksize=kernel_size, strides=pool_strides, padding='SAME') # 定義全連接層函數(shù) def full_connect(inputs, W, b): return tf.nn.relu(tf.matmul(inputs, W) + b) if __name__ =='__main__' and sys.argv[1]=='train': # 第一次遍歷圖片目錄是為了獲取圖片總數(shù) input_count = 0 for i in range(0,NUM_CLASSES): dir = './train_images/training-set/chinese-characters/%s/' % i # 這里可以改成你自己的圖片目錄,i為分類標(biāo)簽 for rt, dirs, files in os.walk(dir): for filename in files: input_count += 1 # 定義對(duì)應(yīng)維數(shù)和各維長(zhǎng)度的數(shù)組 input_images = np.array([[0]*SIZE for i in range(input_count)]) input_labels = np.array([[0]*NUM_CLASSES for i in range(input_count)]) # 第二次遍歷圖片目錄是為了生成圖片數(shù)據(jù)和標(biāo)簽 index = 0 for i in range(0,NUM_CLASSES): dir = './train_images/training-set/chinese-characters/%s/' % i # 這里可以改成你自己的圖片目錄,i為分類標(biāo)簽 for rt, dirs, files in os.walk(dir): for filename in files: filename = dir + filename img = Image.open(filename) width = img.size[0] height = img.size[1] for h in range(0, height): for w in range(0, width): # 通過(guò)這樣的處理,使數(shù)字的線條變細(xì),有利于提高識(shí)別準(zhǔn)確率 if img.getpixel((w, h)) > 230: input_images[index][w+h*width] = 0 else: input_images[index][w+h*width] = 1 input_labels[index][i] = 1 index += 1 # 第一次遍歷圖片目錄是為了獲取圖片總數(shù) val_count = 0 for i in range(0,NUM_CLASSES): dir = './train_images/validation-set/chinese-characters/%s/' % i # 這里可以改成你自己的圖片目錄,i為分類標(biāo)簽 for rt, dirs, files in os.walk(dir): for filename in files: val_count += 1 # 定義對(duì)應(yīng)維數(shù)和各維長(zhǎng)度的數(shù)組 val_images = np.array([[0]*SIZE for i in range(val_count)]) val_labels = np.array([[0]*NUM_CLASSES for i in range(val_count)]) # 第二次遍歷圖片目錄是為了生成圖片數(shù)據(jù)和標(biāo)簽 index = 0 for i in range(0,NUM_CLASSES): dir = './train_images/validation-set/chinese-characters/%s/' % i # 這里可以改成你自己的圖片目錄,i為分類標(biāo)簽 for rt, dirs, files in os.walk(dir): for filename in files: filename = dir + filename img = Image.open(filename) width = img.size[0] height = img.size[1] for h in range(0, height): for w in range(0, width): # 通過(guò)這樣的處理,使數(shù)字的線條變細(xì),有利于提高識(shí)別準(zhǔn)確率 if img.getpixel((w, h)) > 230: val_images[index][w+h*width] = 0 else: val_images[index][w+h*width] = 1 val_labels[index][i] = 1 index += 1 with tf.Session() as sess: # 第一個(gè)卷積層 W_conv1 = tf.Variable(tf.truncated_normal([8, 8, 1, 16], stddev=0.1), name="W_conv1") b_conv1 = tf.Variable(tf.constant(0.1, shape=[16]), name="b_conv1") conv_strides = [1, 1, 1, 1] kernel_size = [1, 2, 2, 1] pool_strides = [1, 2, 2, 1] L1_pool = conv_layer(x_image, W_conv1, b_conv1, conv_strides, kernel_size, pool_strides, padding='SAME') # 第二個(gè)卷積層 W_conv2 = tf.Variable(tf.truncated_normal([5, 5, 16, 32], stddev=0.1), name="W_conv2") b_conv2 = tf.Variable(tf.constant(0.1, shape=[32]), name="b_conv2") conv_strides = [1, 1, 1, 1] kernel_size = [1, 1, 1, 1] pool_strides = [1, 1, 1, 1] L2_pool = conv_layer(L1_pool, W_conv2, b_conv2, conv_strides, kernel_size, pool_strides, padding='SAME') # 全連接層 W_fc1 = tf.Variable(tf.truncated_normal([16 * 20 * 32, 512], stddev=0.1), name="W_fc1") b_fc1 = tf.Variable(tf.constant(0.1, shape=[512]), name="b_fc1") h_pool2_flat = tf.reshape(L2_pool, [-1, 16 * 20*32]) h_fc1 = full_connect(h_pool2_flat, W_fc1, b_fc1) # dropout keep_prob = tf.placeholder(tf.float32) h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob) # readout層 W_fc2 = tf.Variable(tf.truncated_normal([512, NUM_CLASSES], stddev=0.1), name="W_fc2") b_fc2 = tf.Variable(tf.constant(0.1, shape=[NUM_CLASSES]), name="b_fc2") # 定義優(yōu)化器和訓(xùn)練op y_conv = tf.matmul(h_fc1_drop, W_fc2) + b_fc2 cross_entropy = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y_, logits=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)) # 初始化saver saver = tf.train.Saver() sess.run(tf.global_variables_initializer()) time_elapsed = time.time() - time_begin print("讀取圖片文件耗費(fèi)時(shí)間:%d秒" % time_elapsed) time_begin = time.time() print ("一共讀取了 %s 個(gè)訓(xùn)練圖像, %s 個(gè)標(biāo)簽" % (input_count, input_count)) # 設(shè)置每次訓(xùn)練op的輸入個(gè)數(shù)和迭代次數(shù),這里為了支持任意圖片總數(shù),定義了一個(gè)余數(shù)remainder,譬如,如果每次訓(xùn)練op的輸入個(gè)數(shù)為60,圖片總數(shù)為150張,則前面兩次各輸入60張,最后一次輸入30張(余數(shù)30) batch_size = 60 iterations = iterations batches_count = int(input_count / batch_size) remainder = input_count % batch_size print ("訓(xùn)練數(shù)據(jù)集分成 %s 批, 前面每批 %s 個(gè)數(shù)據(jù),最后一批 %s 個(gè)數(shù)據(jù)" % (batches_count+1, batch_size, remainder)) # 執(zhí)行訓(xùn)練迭代 for it in range(iterations): # 這里的關(guān)鍵是要把輸入數(shù)組轉(zhuǎn)為np.array for n in range(batches_count): train_step.run(feed_dict={x: input_images[n*batch_size:(n+1)*batch_size], y_: input_labels[n*batch_size:(n+1)*batch_size], keep_prob: 0.5}) if remainder > 0: start_index = batches_count * batch_size; train_step.run(feed_dict={x: input_images[start_index:input_count-1], y_: input_labels[start_index:input_count-1], keep_prob: 0.5}) # 每完成五次迭代,判斷準(zhǔn)確度是否已達(dá)到100%,達(dá)到則退出迭代循環(huán) iterate_accuracy = 0 if it%5 == 0: iterate_accuracy = accuracy.eval(feed_dict={x: val_images, y_: val_labels, keep_prob: 1.0}) print ('第 %d 次訓(xùn)練迭代: 準(zhǔn)確率 %0.5f%%' % (it, iterate_accuracy*100)) if iterate_accuracy >= 0.9999 and it >= 150: break; print ('完成訓(xùn)練!') time_elapsed = time.time() - time_begin print ("訓(xùn)練耗費(fèi)時(shí)間:%d秒" % time_elapsed) time_begin = time.time() # 保存訓(xùn)練結(jié)果 if not os.path.exists(SAVER_DIR): print ('不存在訓(xùn)練數(shù)據(jù)保存目錄,現(xiàn)在創(chuàng)建保存目錄') os.makedirs(SAVER_DIR) saver_path = saver.save(sess, "%smodel.ckpt"%(SAVER_DIR)) if __name__ =='__main__' and sys.argv[1]=='predict': saver = tf.train.import_meta_graph("%smodel.ckpt.meta"%(SAVER_DIR)) with tf.Session() as sess: model_file=tf.train.latest_checkpoint(SAVER_DIR) saver.restore(sess, model_file) # 第一個(gè)卷積層 W_conv1 = sess.graph.get_tensor_by_name("W_conv1:0") b_conv1 = sess.graph.get_tensor_by_name("b_conv1:0") conv_strides = [1, 1, 1, 1] kernel_size = [1, 2, 2, 1] pool_strides = [1, 2, 2, 1] L1_pool = conv_layer(x_image, W_conv1, b_conv1, conv_strides, kernel_size, pool_strides, padding='SAME') # 第二個(gè)卷積層 W_conv2 = sess.graph.get_tensor_by_name("W_conv2:0") b_conv2 = sess.graph.get_tensor_by_name("b_conv2:0") conv_strides = [1, 1, 1, 1] kernel_size = [1, 1, 1, 1] pool_strides = [1, 1, 1, 1] L2_pool = conv_layer(L1_pool, W_conv2, b_conv2, conv_strides, kernel_size, pool_strides, padding='SAME') # 全連接層 W_fc1 = sess.graph.get_tensor_by_name("W_fc1:0") b_fc1 = sess.graph.get_tensor_by_name("b_fc1:0") h_pool2_flat = tf.reshape(L2_pool, [-1, 16 * 20*32]) h_fc1 = full_connect(h_pool2_flat, W_fc1, b_fc1) # dropout keep_prob = tf.placeholder(tf.float32) h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob) # readout層 W_fc2 = sess.graph.get_tensor_by_name("W_fc2:0") b_fc2 = sess.graph.get_tensor_by_name("b_fc2:0") # 定義優(yōu)化器和訓(xùn)練op conv = tf.nn.softmax(tf.matmul(h_fc1_drop, W_fc2) + b_fc2) for n in range(1,2): path = "test_images/%s.bmp" % (n) img = Image.open(path) width = img.size[0] height = img.size[1] img_data = [[0]*SIZE for i in range(1)] for h in range(0, height): for w in range(0, width): if img.getpixel((w, h)) < 190: img_data[0][w+h*width] = 1 else: img_data[0][w+h*width] = 0 result = sess.run(conv, feed_dict = {x: np.array(img_data), keep_prob: 1.0}) max1 = 0 max2 = 0 max3 = 0 max1_index = 0 max2_index = 0 max3_index = 0 for j in range(NUM_CLASSES): if result[0][j] > max1: max1 = result[0][j] max1_index = j continue if (result[0][j]>max2) and (result[0][j]<=max1): max2 = result[0][j] max2_index = j continue if (result[0][j]>max3) and (result[0][j]<=max2): max3 = result[0][j] max3_index = j continue nProvinceIndex = max1_index print ("概率: [%s %0.2f%%] [%s %0.2f%%] [%s %0.2f%%]" % (PROVINCES[max1_index],max1*100, PROVINCES[max2_index],max2*100, PROVINCES[max3_index],max3*100)) print ("省份簡(jiǎn)稱是: %s" % PROVINCES[nProvinceIndex])
城市代號(hào)訓(xùn)練+識(shí)別代碼(保存文件名為train-license-letters.py):
#!/usr/bin/python3.5 # -*- coding: utf-8 -*- import sys import os import time import random import numpy as np import tensorflow as tf from PIL import Image SIZE = 1280 WIDTH = 32 HEIGHT = 40 NUM_CLASSES = 26 iterations = 500 SAVER_DIR = "train-saver/letters/" LETTERS_DIGITS = ("A","B","C","D","E","F","G","H","J","K","L","M","N","P","Q","R","S","T","U","V","W","X","Y","Z","I","O") license_num = "" time_begin = time.time() # 定義輸入節(jié)點(diǎn),對(duì)應(yīng)于圖片像素值矩陣集合和圖片標(biāo)簽(即所代表的數(shù)字) x = tf.placeholder(tf.float32, shape=[None, SIZE]) y_ = tf.placeholder(tf.float32, shape=[None, NUM_CLASSES]) x_image = tf.reshape(x, [-1, WIDTH, HEIGHT, 1]) # 定義卷積函數(shù) def conv_layer(inputs, W, b, conv_strides, kernel_size, pool_strides, padding): L1_conv = tf.nn.conv2d(inputs, W, strides=conv_strides, padding=padding) L1_relu = tf.nn.relu(L1_conv + b) return tf.nn.max_pool(L1_relu, ksize=kernel_size, strides=pool_strides, padding='SAME') # 定義全連接層函數(shù) def full_connect(inputs, W, b): return tf.nn.relu(tf.matmul(inputs, W) + b) if __name__ =='__main__' and sys.argv[1]=='train': # 第一次遍歷圖片目錄是為了獲取圖片總數(shù) input_count = 0 for i in range(0+10,NUM_CLASSES+10): dir = './train_images/training-set/letters/%s/' % i # 這里可以改成你自己的圖片目錄,i為分類標(biāo)簽 for rt, dirs, files in os.walk(dir): for filename in files: input_count += 1 # 定義對(duì)應(yīng)維數(shù)和各維長(zhǎng)度的數(shù)組 input_images = np.array([[0]*SIZE for i in range(input_count)]) input_labels = np.array([[0]*NUM_CLASSES for i in range(input_count)]) # 第二次遍歷圖片目錄是為了生成圖片數(shù)據(jù)和標(biāo)簽 index = 0 for i in range(0+10,NUM_CLASSES+10): dir = './train_images/training-set/letters/%s/' % i # 這里可以改成你自己的圖片目錄,i為分類標(biāo)簽 for rt, dirs, files in os.walk(dir): for filename in files: filename = dir + filename img = Image.open(filename) width = img.size[0] height = img.size[1] for h in range(0, height): for w in range(0, width): # 通過(guò)這樣的處理,使數(shù)字的線條變細(xì),有利于提高識(shí)別準(zhǔn)確率 if img.getpixel((w, h)) > 230: input_images[index][w+h*width] = 0 else: input_images[index][w+h*width] = 1 #print ("i=%d, index=%d" % (i, index)) input_labels[index][i-10] = 1 index += 1 # 第一次遍歷圖片目錄是為了獲取圖片總數(shù) val_count = 0 for i in range(0+10,NUM_CLASSES+10): dir = './train_images/validation-set/%s/' % i # 這里可以改成你自己的圖片目錄,i為分類標(biāo)簽 for rt, dirs, files in os.walk(dir): for filename in files: val_count += 1 # 定義對(duì)應(yīng)維數(shù)和各維長(zhǎng)度的數(shù)組 val_images = np.array([[0]*SIZE for i in range(val_count)]) val_labels = np.array([[0]*NUM_CLASSES for i in range(val_count)]) # 第二次遍歷圖片目錄是為了生成圖片數(shù)據(jù)和標(biāo)簽 index = 0 for i in range(0+10,NUM_CLASSES+10): dir = './train_images/validation-set/%s/' % i # 這里可以改成你自己的圖片目錄,i為分類標(biāo)簽 for rt, dirs, files in os.walk(dir): for filename in files: filename = dir + filename img = Image.open(filename) width = img.size[0] height = img.size[1] for h in range(0, height): for w in range(0, width): # 通過(guò)這樣的處理,使數(shù)字的線條變細(xì),有利于提高識(shí)別準(zhǔn)確率 if img.getpixel((w, h)) > 230: val_images[index][w+h*width] = 0 else: val_images[index][w+h*width] = 1 val_labels[index][i-10] = 1 index += 1 with tf.Session() as sess: # 第一個(gè)卷積層 W_conv1 = tf.Variable(tf.truncated_normal([8, 8, 1, 16], stddev=0.1), name="W_conv1") b_conv1 = tf.Variable(tf.constant(0.1, shape=[16]), name="b_conv1") conv_strides = [1, 1, 1, 1] kernel_size = [1, 2, 2, 1] pool_strides = [1, 2, 2, 1] L1_pool = conv_layer(x_image, W_conv1, b_conv1, conv_strides, kernel_size, pool_strides, padding='SAME') # 第二個(gè)卷積層 W_conv2 = tf.Variable(tf.truncated_normal([5, 5, 16, 32], stddev=0.1), name="W_conv2") b_conv2 = tf.Variable(tf.constant(0.1, shape=[32]), name="b_conv2") conv_strides = [1, 1, 1, 1] kernel_size = [1, 1, 1, 1] pool_strides = [1, 1, 1, 1] L2_pool = conv_layer(L1_pool, W_conv2, b_conv2, conv_strides, kernel_size, pool_strides, padding='SAME') # 全連接層 W_fc1 = tf.Variable(tf.truncated_normal([16 * 20 * 32, 512], stddev=0.1), name="W_fc1") b_fc1 = tf.Variable(tf.constant(0.1, shape=[512]), name="b_fc1") h_pool2_flat = tf.reshape(L2_pool, [-1, 16 * 20*32]) h_fc1 = full_connect(h_pool2_flat, W_fc1, b_fc1) # dropout keep_prob = tf.placeholder(tf.float32) h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob) # readout層 W_fc2 = tf.Variable(tf.truncated_normal([512, NUM_CLASSES], stddev=0.1), name="W_fc2") b_fc2 = tf.Variable(tf.constant(0.1, shape=[NUM_CLASSES]), name="b_fc2") # 定義優(yōu)化器和訓(xùn)練op y_conv = tf.matmul(h_fc1_drop, W_fc2) + b_fc2 cross_entropy = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y_, logits=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.global_variables_initializer()) time_elapsed = time.time() - time_begin print("讀取圖片文件耗費(fèi)時(shí)間:%d秒" % time_elapsed) time_begin = time.time() print ("一共讀取了 %s 個(gè)訓(xùn)練圖像, %s 個(gè)標(biāo)簽" % (input_count, input_count)) # 設(shè)置每次訓(xùn)練op的輸入個(gè)數(shù)和迭代次數(shù),這里為了支持任意圖片總數(shù),定義了一個(gè)余數(shù)remainder,譬如,如果每次訓(xùn)練op的輸入個(gè)數(shù)為60,圖片總數(shù)為150張,則前面兩次各輸入60張,最后一次輸入30張(余數(shù)30) batch_size = 60 iterations = iterations batches_count = int(input_count / batch_size) remainder = input_count % batch_size print ("訓(xùn)練數(shù)據(jù)集分成 %s 批, 前面每批 %s 個(gè)數(shù)據(jù),最后一批 %s 個(gè)數(shù)據(jù)" % (batches_count+1, batch_size, remainder)) # 執(zhí)行訓(xùn)練迭代 for it in range(iterations): # 這里的關(guān)鍵是要把輸入數(shù)組轉(zhuǎn)為np.array for n in range(batches_count): train_step.run(feed_dict={x: input_images[n*batch_size:(n+1)*batch_size], y_: input_labels[n*batch_size:(n+1)*batch_size], keep_prob: 0.5}) if remainder > 0: start_index = batches_count * batch_size; train_step.run(feed_dict={x: input_images[start_index:input_count-1], y_: input_labels[start_index:input_count-1], keep_prob: 0.5}) # 每完成五次迭代,判斷準(zhǔn)確度是否已達(dá)到100%,達(dá)到則退出迭代循環(huán) iterate_accuracy = 0 if it%5 == 0: iterate_accuracy = accuracy.eval(feed_dict={x: val_images, y_: val_labels, keep_prob: 1.0}) print ('第 %d 次訓(xùn)練迭代: 準(zhǔn)確率 %0.5f%%' % (it, iterate_accuracy*100)) if iterate_accuracy >= 0.9999 and it >= iterations: break; print ('完成訓(xùn)練!') time_elapsed = time.time() - time_begin print ("訓(xùn)練耗費(fèi)時(shí)間:%d秒" % time_elapsed) time_begin = time.time() # 保存訓(xùn)練結(jié)果 if not os.path.exists(SAVER_DIR): print ('不存在訓(xùn)練數(shù)據(jù)保存目錄,現(xiàn)在創(chuàng)建保存目錄') os.makedirs(SAVER_DIR) # 初始化saver saver = tf.train.Saver() saver_path = saver.save(sess, "%smodel.ckpt"%(SAVER_DIR)) if __name__ =='__main__' and sys.argv[1]=='predict': saver = tf.train.import_meta_graph("%smodel.ckpt.meta"%(SAVER_DIR)) with tf.Session() as sess: model_file=tf.train.latest_checkpoint(SAVER_DIR) saver.restore(sess, model_file) # 第一個(gè)卷積層 W_conv1 = sess.graph.get_tensor_by_name("W_conv1:0") b_conv1 = sess.graph.get_tensor_by_name("b_conv1:0") conv_strides = [1, 1, 1, 1] kernel_size = [1, 2, 2, 1] pool_strides = [1, 2, 2, 1] L1_pool = conv_layer(x_image, W_conv1, b_conv1, conv_strides, kernel_size, pool_strides, padding='SAME') # 第二個(gè)卷積層 W_conv2 = sess.graph.get_tensor_by_name("W_conv2:0") b_conv2 = sess.graph.get_tensor_by_name("b_conv2:0") conv_strides = [1, 1, 1, 1] kernel_size = [1, 1, 1, 1] pool_strides = [1, 1, 1, 1] L2_pool = conv_layer(L1_pool, W_conv2, b_conv2, conv_strides, kernel_size, pool_strides, padding='SAME') # 全連接層 W_fc1 = sess.graph.get_tensor_by_name("W_fc1:0") b_fc1 = sess.graph.get_tensor_by_name("b_fc1:0") h_pool2_flat = tf.reshape(L2_pool, [-1, 16 * 20*32]) h_fc1 = full_connect(h_pool2_flat, W_fc1, b_fc1) # dropout keep_prob = tf.placeholder(tf.float32) h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob) # readout層 W_fc2 = sess.graph.get_tensor_by_name("W_fc2:0") b_fc2 = sess.graph.get_tensor_by_name("b_fc2:0") # 定義優(yōu)化器和訓(xùn)練op conv = tf.nn.softmax(tf.matmul(h_fc1_drop, W_fc2) + b_fc2) for n in range(2,3): path = "test_images/%s.bmp" % (n) img = Image.open(path) width = img.size[0] height = img.size[1] img_data = [[0]*SIZE for i in range(1)] for h in range(0, height): for w in range(0, width): if img.getpixel((w, h)) < 190: img_data[0][w+h*width] = 1 else: img_data[0][w+h*width] = 0 result = sess.run(conv, feed_dict = {x: np.array(img_data), keep_prob: 1.0}) max1 = 0 max2 = 0 max3 = 0 max1_index = 0 max2_index = 0 max3_index = 0 for j in range(NUM_CLASSES): if result[0][j] > max1: max1 = result[0][j] max1_index = j continue if (result[0][j]>max2) and (result[0][j]<=max1): max2 = result[0][j] max2_index = j continue if (result[0][j]>max3) and (result[0][j]<=max2): max3 = result[0][j] max3_index = j continue if n == 3: license_num += "-" license_num = license_num + LETTERS_DIGITS[max1_index] print ("概率: [%s %0.2f%%] [%s %0.2f%%] [%s %0.2f%%]" % (LETTERS_DIGITS[max1_index],max1*100, LETTERS_DIGITS[max2_index],max2*100, LETTERS_DIGITS[max3_index],max3*100)) print ("城市代號(hào)是: 【%s】" % license_num)
車牌編號(hào)訓(xùn)練+識(shí)別代碼(保存文件名為train-license-digits.py):
#!/usr/bin/python3.5 # -*- coding: utf-8 -*- import sys import os import time import random import numpy as np import tensorflow as tf from PIL import Image SIZE = 1280 WIDTH = 32 HEIGHT = 40 NUM_CLASSES = 34 iterations = 1000 SAVER_DIR = "train-saver/digits/" LETTERS_DIGITS = ("0","1","2","3","4","5","6","7","8","9","A","B","C","D","E","F","G","H","J","K","L","M","N","P","Q","R","S","T","U","V","W","X","Y","Z") license_num = "" time_begin = time.time() # 定義輸入節(jié)點(diǎn),對(duì)應(yīng)于圖片像素值矩陣集合和圖片標(biāo)簽(即所代表的數(shù)字) x = tf.placeholder(tf.float32, shape=[None, SIZE]) y_ = tf.placeholder(tf.float32, shape=[None, NUM_CLASSES]) x_image = tf.reshape(x, [-1, WIDTH, HEIGHT, 1]) # 定義卷積函數(shù) def conv_layer(inputs, W, b, conv_strides, kernel_size, pool_strides, padding): L1_conv = tf.nn.conv2d(inputs, W, strides=conv_strides, padding=padding) L1_relu = tf.nn.relu(L1_conv + b) return tf.nn.max_pool(L1_relu, ksize=kernel_size, strides=pool_strides, padding='SAME') # 定義全連接層函數(shù) def full_connect(inputs, W, b): return tf.nn.relu(tf.matmul(inputs, W) + b) if __name__ =='__main__' and sys.argv[1]=='train': # 第一次遍歷圖片目錄是為了獲取圖片總數(shù) input_count = 0 for i in range(0,NUM_CLASSES): dir = './train_images/training-set/%s/' % i # 這里可以改成你自己的圖片目錄,i為分類標(biāo)簽 for rt, dirs, files in os.walk(dir): for filename in files: input_count += 1 # 定義對(duì)應(yīng)維數(shù)和各維長(zhǎng)度的數(shù)組 input_images = np.array([[0]*SIZE for i in range(input_count)]) input_labels = np.array([[0]*NUM_CLASSES for i in range(input_count)]) # 第二次遍歷圖片目錄是為了生成圖片數(shù)據(jù)和標(biāo)簽 index = 0 for i in range(0,NUM_CLASSES): dir = './train_images/training-set/%s/' % i # 這里可以改成你自己的圖片目錄,i為分類標(biāo)簽 for rt, dirs, files in os.walk(dir): for filename in files: filename = dir + filename img = Image.open(filename) width = img.size[0] height = img.size[1] for h in range(0, height): for w in range(0, width): # 通過(guò)這樣的處理,使數(shù)字的線條變細(xì),有利于提高識(shí)別準(zhǔn)確率 if img.getpixel((w, h)) > 230: input_images[index][w+h*width] = 0 else: input_images[index][w+h*width] = 1 input_labels[index][i] = 1 index += 1 # 第一次遍歷圖片目錄是為了獲取圖片總數(shù) val_count = 0 for i in range(0,NUM_CLASSES): dir = './train_images/validation-set/%s/' % i # 這里可以改成你自己的圖片目錄,i為分類標(biāo)簽 for rt, dirs, files in os.walk(dir): for filename in files: val_count += 1 # 定義對(duì)應(yīng)維數(shù)和各維長(zhǎng)度的數(shù)組 val_images = np.array([[0]*SIZE for i in range(val_count)]) val_labels = np.array([[0]*NUM_CLASSES for i in range(val_count)]) # 第二次遍歷圖片目錄是為了生成圖片數(shù)據(jù)和標(biāo)簽 index = 0 for i in range(0,NUM_CLASSES): dir = './train_images/validation-set/%s/' % i # 這里可以改成你自己的圖片目錄,i為分類標(biāo)簽 for rt, dirs, files in os.walk(dir): for filename in files: filename = dir + filename img = Image.open(filename) width = img.size[0] height = img.size[1] for h in range(0, height): for w in range(0, width): # 通過(guò)這樣的處理,使數(shù)字的線條變細(xì),有利于提高識(shí)別準(zhǔn)確率 if img.getpixel((w, h)) > 230: val_images[index][w+h*width] = 0 else: val_images[index][w+h*width] = 1 val_labels[index][i] = 1 index += 1 with tf.Session() as sess: # 第一個(gè)卷積層 W_conv1 = tf.Variable(tf.truncated_normal([8, 8, 1, 16], stddev=0.1), name="W_conv1") b_conv1 = tf.Variable(tf.constant(0.1, shape=[16]), name="b_conv1") conv_strides = [1, 1, 1, 1] kernel_size = [1, 2, 2, 1] pool_strides = [1, 2, 2, 1] L1_pool = conv_layer(x_image, W_conv1, b_conv1, conv_strides, kernel_size, pool_strides, padding='SAME') # 第二個(gè)卷積層 W_conv2 = tf.Variable(tf.truncated_normal([5, 5, 16, 32], stddev=0.1), name="W_conv2") b_conv2 = tf.Variable(tf.constant(0.1, shape=[32]), name="b_conv2") conv_strides = [1, 1, 1, 1] kernel_size = [1, 1, 1, 1] pool_strides = [1, 1, 1, 1] L2_pool = conv_layer(L1_pool, W_conv2, b_conv2, conv_strides, kernel_size, pool_strides, padding='SAME') # 全連接層 W_fc1 = tf.Variable(tf.truncated_normal([16 * 20 * 32, 512], stddev=0.1), name="W_fc1") b_fc1 = tf.Variable(tf.constant(0.1, shape=[512]), name="b_fc1") h_pool2_flat = tf.reshape(L2_pool, [-1, 16 * 20*32]) h_fc1 = full_connect(h_pool2_flat, W_fc1, b_fc1) # dropout keep_prob = tf.placeholder(tf.float32) h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob) # readout層 W_fc2 = tf.Variable(tf.truncated_normal([512, NUM_CLASSES], stddev=0.1), name="W_fc2") b_fc2 = tf.Variable(tf.constant(0.1, shape=[NUM_CLASSES]), name="b_fc2") # 定義優(yōu)化器和訓(xùn)練op y_conv = tf.matmul(h_fc1_drop, W_fc2) + b_fc2 cross_entropy = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y_, logits=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.global_variables_initializer()) time_elapsed = time.time() - time_begin print("讀取圖片文件耗費(fèi)時(shí)間:%d秒" % time_elapsed) time_begin = time.time() print ("一共讀取了 %s 個(gè)訓(xùn)練圖像, %s 個(gè)標(biāo)簽" % (input_count, input_count)) # 設(shè)置每次訓(xùn)練op的輸入個(gè)數(shù)和迭代次數(shù),這里為了支持任意圖片總數(shù),定義了一個(gè)余數(shù)remainder,譬如,如果每次訓(xùn)練op的輸入個(gè)數(shù)為60,圖片總數(shù)為150張,則前面兩次各輸入60張,最后一次輸入30張(余數(shù)30) batch_size = 60 iterations = iterations batches_count = int(input_count / batch_size) remainder = input_count % batch_size print ("訓(xùn)練數(shù)據(jù)集分成 %s 批, 前面每批 %s 個(gè)數(shù)據(jù),最后一批 %s 個(gè)數(shù)據(jù)" % (batches_count+1, batch_size, remainder)) # 執(zhí)行訓(xùn)練迭代 for it in range(iterations): # 這里的關(guān)鍵是要把輸入數(shù)組轉(zhuǎn)為np.array for n in range(batches_count): train_step.run(feed_dict={x: input_images[n*batch_size:(n+1)*batch_size], y_: input_labels[n*batch_size:(n+1)*batch_size], keep_prob: 0.5}) if remainder > 0: start_index = batches_count * batch_size; train_step.run(feed_dict={x: input_images[start_index:input_count-1], y_: input_labels[start_index:input_count-1], keep_prob: 0.5}) # 每完成五次迭代,判斷準(zhǔn)確度是否已達(dá)到100%,達(dá)到則退出迭代循環(huán) iterate_accuracy = 0 if it%5 == 0: iterate_accuracy = accuracy.eval(feed_dict={x: val_images, y_: val_labels, keep_prob: 1.0}) print ('第 %d 次訓(xùn)練迭代: 準(zhǔn)確率 %0.5f%%' % (it, iterate_accuracy*100)) if iterate_accuracy >= 0.9999 and it >= iterations: break; print ('完成訓(xùn)練!') time_elapsed = time.time() - time_begin print ("訓(xùn)練耗費(fèi)時(shí)間:%d秒" % time_elapsed) time_begin = time.time() # 保存訓(xùn)練結(jié)果 if not os.path.exists(SAVER_DIR): print ('不存在訓(xùn)練數(shù)據(jù)保存目錄,現(xiàn)在創(chuàng)建保存目錄') os.makedirs(SAVER_DIR) # 初始化saver saver = tf.train.Saver() saver_path = saver.save(sess, "%smodel.ckpt"%(SAVER_DIR)) if __name__ =='__main__' and sys.argv[1]=='predict': saver = tf.train.import_meta_graph("%smodel.ckpt.meta"%(SAVER_DIR)) with tf.Session() as sess: model_file=tf.train.latest_checkpoint(SAVER_DIR) saver.restore(sess, model_file) # 第一個(gè)卷積層 W_conv1 = sess.graph.get_tensor_by_name("W_conv1:0") b_conv1 = sess.graph.get_tensor_by_name("b_conv1:0") conv_strides = [1, 1, 1, 1] kernel_size = [1, 2, 2, 1] pool_strides = [1, 2, 2, 1] L1_pool = conv_layer(x_image, W_conv1, b_conv1, conv_strides, kernel_size, pool_strides, padding='SAME') # 第二個(gè)卷積層 W_conv2 = sess.graph.get_tensor_by_name("W_conv2:0") b_conv2 = sess.graph.get_tensor_by_name("b_conv2:0") conv_strides = [1, 1, 1, 1] kernel_size = [1, 1, 1, 1] pool_strides = [1, 1, 1, 1] L2_pool = conv_layer(L1_pool, W_conv2, b_conv2, conv_strides, kernel_size, pool_strides, padding='SAME') # 全連接層 W_fc1 = sess.graph.get_tensor_by_name("W_fc1:0") b_fc1 = sess.graph.get_tensor_by_name("b_fc1:0") h_pool2_flat = tf.reshape(L2_pool, [-1, 16 * 20*32]) h_fc1 = full_connect(h_pool2_flat, W_fc1, b_fc1) # dropout keep_prob = tf.placeholder(tf.float32) h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob) # readout層 W_fc2 = sess.graph.get_tensor_by_name("W_fc2:0") b_fc2 = sess.graph.get_tensor_by_name("b_fc2:0") # 定義優(yōu)化器和訓(xùn)練op conv = tf.nn.softmax(tf.matmul(h_fc1_drop, W_fc2) + b_fc2) for n in range(3,8): path = "test_images/%s.bmp" % (n) img = Image.open(path) width = img.size[0] height = img.size[1] img_data = [[0]*SIZE for i in range(1)] for h in range(0, height): for w in range(0, width): if img.getpixel((w, h)) < 190: img_data[0][w+h*width] = 1 else: img_data[0][w+h*width] = 0 result = sess.run(conv, feed_dict = {x: np.array(img_data), keep_prob: 1.0}) max1 = 0 max2 = 0 max3 = 0 max1_index = 0 max2_index = 0 max3_index = 0 for j in range(NUM_CLASSES): if result[0][j] > max1: max1 = result[0][j] max1_index = j continue if (result[0][j]>max2) and (result[0][j]<=max1): max2 = result[0][j] max2_index = j continue if (result[0][j]>max3) and (result[0][j]<=max2): max3 = result[0][j] max3_index = j continue license_num = license_num + LETTERS_DIGITS[max1_index] print ("概率: [%s %0.2f%%] [%s %0.2f%%] [%s %0.2f%%]" % (LETTERS_DIGITS[max1_index],max1*100, LETTERS_DIGITS[max2_index],max2*100, LETTERS_DIGITS[max3_index],max3*100)) print ("車牌編號(hào)是: 【%s】" % license_num)
保存好上面三個(gè)python腳本后,我們首先進(jìn)行省份簡(jiǎn)稱訓(xùn)練。在運(yùn)行代碼之前,需要先把數(shù)據(jù)集解壓到訓(xùn)練腳本所在目錄。然后,在命令行中進(jìn)入腳本所在目錄,輸入執(zhí)行如下命令:
python train-license-province.py train
訓(xùn)練結(jié)果如下:
然后進(jìn)行省份簡(jiǎn)稱識(shí)別,在命令行輸入執(zhí)行如下命令:
python train-license-province.py predict
執(zhí)行城市代號(hào)訓(xùn)練(相當(dāng)于訓(xùn)練26個(gè)字母):
python train-license-letters.py train
識(shí)別城市代號(hào):
python train-license-letters.py predict
執(zhí)行車牌編號(hào)訓(xùn)練(相當(dāng)于訓(xùn)練24個(gè)字母+10個(gè)數(shù)字,我國(guó)交通法規(guī)規(guī)定車牌編號(hào)中不包含字母I和O):
python train-license-digits.py train
識(shí)別車牌編號(hào):
python train-license-digits.py predict
可以看到,在測(cè)試圖片上,識(shí)別準(zhǔn)確率很高。識(shí)別結(jié)果是閩O-1672Q。
下圖是測(cè)試圖片的車牌原圖:
以上是“TensorFlow如何實(shí)現(xiàn)車牌識(shí)別功能”這篇文章的所有內(nèi)容,感謝各位的閱讀!相信大家都有了一定的了解,希望分享的內(nèi)容對(duì)大家有所幫助,如果還想學(xué)習(xí)更多知識(shí),歡迎關(guān)注億速云行業(yè)資訊頻道!
免責(zé)聲明:本站發(fā)布的內(nèi)容(圖片、視頻和文字)以原創(chuàng)、轉(zhuǎn)載和分享為主,文章觀點(diǎn)不代表本網(wǎng)站立場(chǎng),如果涉及侵權(quán)請(qǐng)聯(lián)系站長(zhǎng)郵箱:is@yisu.com進(jìn)行舉報(bào),并提供相關(guān)證據(jù),一經(jīng)查實(shí),將立刻刪除涉嫌侵權(quán)內(nèi)容。