TensorFlow车牌识别完整版（含车牌数据集）
在之前发布的⼀篇博⽂《》中，我们演⽰了如何使⽤TensorFlow进⾏车牌识别，但是，当时采⽤的数据集是MNIST数字⼿写体，只能分类0-9共10个数字，⽆法分类省份简称和字母，局限性较⼤，⽆实际意义。
经过图像定位分割处理，博主收集了相关省份简称和26个字母的图⽚数据集，结合前述博⽂中贴出的python+TensorFlow代码，实现了完整的车牌识别功能。本着分享精神，在此送上全部代码和车牌数据集。
省份简称训练+识别代码（保存⽂件名为train-license-province.py）（拷贝代码请务必注意python⽂本缩进，只要有⼀处缩进错误，就⽆法得到正确结果，或者出现异常）：
#!/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","闽","粤","苏","沪","浙")
nProvinceIndex = 0
time_begin = time.time()
# 定义输⼊节点，对应于图⽚像素值矩阵集合和图⽚标签(即所代表的数字)
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])
# 定义卷积函数
def conv_layer(inputs, W, b, conv_strides, kernel_size, pool_strides, padding):
L1_conv = v2d(inputs, W, strides=conv_strides, padding=padding)
L1_relu = lu(L1_conv + b)
ax_pool(L1_relu, ksize=kernel_size, strides=pool_strides, padding='SAME')
# 定义全连接层函数
def full_connect(inputs, W, b):
lu(tf.matmul(inputs, W) + b)
if __name__ =='__main__' and sys.argv[1]=='train':
# 第⼀次遍历图⽚⽬录是为了获取图⽚总数
input_count = 0
for i in range(0,NUM_CLASSES):
dir = './train_images/training-set/chinese-characters/%s/' % i          # 这⾥可以改成你⾃⼰的图⽚⽬录，i为分类标签
for rt, dirs, files in os.walk(dir):
for filename in files:
for filename in files:
input_count += 1
# 定义对应维数和各维长度的数组
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)])
# 第⼆次遍历图⽚⽬录是为了⽣成图⽚数据和标签
index = 0
for i in range(0,NUM_CLASSES):
dir = './train_images/training-set/chinese-characters/%s/' % i          # 这⾥可以改成你⾃⼰的图⽚⽬录，i为分类标签        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):
# 通过这样的处理，使数字的线条变细，有利于提⾼识别准确率
session如何设置和读取pixel((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
# 第⼀次遍历图⽚⽬录是为了获取图⽚总数
val_count = 0
for i in range(0,NUM_CLASSES):
dir = './train_images/validation-set/chinese-characters/%s/' % i          # 这⾥可以改成你⾃⼰的图⽚⽬录，i为分类标签        for rt, dirs, files in os.walk(dir):
for filename in files:
val_count += 1
# 定义对应维数和各维长度的数组
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)])
# 第⼆次遍历图⽚⽬录是为了⽣成图⽚数据和标签
index = 0
for i in range(0,NUM_CLASSES):
dir = './train_images/validation-set/chinese-characters/%s/' % i          # 这⾥可以改成你⾃⼰的图⽚⽬录，i为分类标签        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):
# 通过这样的处理，使数字的线条变细，有利于提⾼识别准确率
pixel((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:
# 第⼀个卷积层
W_conv1 = tf.uncated_normal([8, 8, 1, 16], stddev=0.1), name="W_conv1")
b_conv1 = tf.stant(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]
pool_strides = [1, 2, 2, 1]
L1_pool = conv_layer(x_image, W_conv1, b_conv1, conv_strides, kernel_size, pool_strides, padding='SAME')
# 第⼆个卷积层
W_conv2 = tf.uncated_normal([5, 5, 16, 32], stddev=0.1), name="W_conv2")
b_conv2 = tf.stant(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')
# 全连接层
W_fc1 = tf.uncated_normal([16 * 20 * 32, 512], stddev=0.1), name="W_fc1")
b_fc1 = tf.stant(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层
W_fc2 = tf.uncated_normal([512, NUM_CLASSES], stddev=0.1), name="W_fc2")
b_fc2 = tf.stant(0.1, shape=[NUM_CLASSES]), name="b_fc2")
# 定义优化器和训练op
y_conv = tf.matmul(h_fc1_drop, W_fc2) + b_fc2
cross_entropy = tf.reduce_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("读取图⽚⽂件耗费时间：%d秒" % time_elapsed)
time_begin = time.time()
print ("⼀共读取了 %s 个训练图像， %s 个标签" % (input_count, input_count))
# 设置每次训练op的输⼊个数和迭代次数，这⾥为了⽀持任意图⽚总数，定义了⼀个余数remainder，譬如，如果每次训练op的输⼊个数为60，图⽚总数为150张        batch_size = 60
iterations = iterations
batches_count = int(input_count / batch_size)
remainder = input_count % batch_size
print ("训练数据集分成 %s 批, 前⾯每批 %s 个数据，最后⼀批 %s 个数据" % (batches_count+1, batch_size, remainder))
# 执⾏训练迭代
for it in range(iterations):
# 这⾥的关键是要把输⼊数组转为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，判断准确度是否已达到100%，达到则退出迭代循环
iterate_accuracy = 0
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 次训练迭代: 准确率 %0.5f%%' % (it, iterate_accuracy*100))
if iterate_accuracy >= 0.9999 and it >= 150:
break;
print ('完成训练!')
time_elapsed = time.time() - time_begin
print ("训练耗费时间：%d秒" % time_elapsed)
time_begin = time.time()
# 保存训练结果
if not ists(SAVER_DIR):
print ('不存在训练数据保存⽬录，现在创建保存⽬录')
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("%a"%(SAVER_DIR))
with tf.Session() as sess:
model_ain.latest_checkpoint(SAVER_DIR)
# 第⼀个卷积层
W_conv1 = _tensor_by_name("W_conv1:0")
b_conv1 = _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')
# 第⼆个卷积层
W_conv2 = _tensor_by_name("W_conv2:0")
b_conv2 = _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')
# 全连接层
W_fc1 = _tensor_by_name("W_fc1:0")
b_fc1 = _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层
W_fc2 = _tensor_by_name("W_fc2:0")
b_fc2 = _tensor_by_name("b_fc2:0")
# 定义优化器和训练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)
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):
pixel((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, PROVINC        print ("省份简称是: %s" % PROVINCES[nProvinceIndex])
城市代号训练+识别代码（保存⽂件名为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")

版权声明：本站内容均来自互联网，仅供演示用，请勿用于商业和其他非法用途。如果侵犯了您的权益请与我们联系QQ:729038198，我们将在24小时内删除。