基于Tensorflow实现AlexNet，实现大陆车牌单个字符的65分类，提供训练和测试数据集的tfrecord文件

import tensorflow as tf import numpy as np from PIL import Image import inference import train import math import time import os os.environ["CUDA_VISIBLE_DEVICES"] = "1" e = math.e # 接收的两个参数 # image_path = '' # model = 0 #start = time.time() def recognistion(image_path): tf.reset_default_graph() # 根据model参数值的不同来选择不同的模型 # 得到测试图片 image_raw_data = tf.gfile.FastGFile(image_path, 'rb').read() x = tf.image.decode_jpeg(image_raw_data) x = tf.cast(x, tf.float32) x = tf.reshape(x, [-1, 32, 16, 1]) # image = Image.open(image_path) # image = tf.cast(image,tf.float32) # x = tf.reshape(image,[32,16,1]) # x = np.asarray(image.convert('L')) # x = np.asarray(Image.open(image_path).convert('L')) # 测试图片的前向传播过程 y = inference.inference(x, None, None) # y_ = tf.array(y) # p = np.amax(y_, axis=0) # 得到置信度最高项的下标 prediction = tf.argmax(y, 1) # 通过变量重命名的方式加载模型 variable_averages = tf.train.ExponentialMovingAverage(train.MOVINGING_AVERAGE_DECAY) variables_to_restore = variable_averages.variables_to_restore() saver = tf.train.Saver(variables_to_restore) with tf.Session() as sess: sess.run((tf.global_variables_initializer(), tf.local_variables_initializer())) # tf.train.get_checkpoint_state函数会通过checkpoint文件自动找到目录中最新模型的文件 ckpt = tf.train.get_checkpoint_state(train.MODEL_SAVE_PATH) if ckpt and ckpt.model_checkpoint_path: # 加载模型 saver.restore(sess, ckpt.model_checkpoint_path) # 通过文件名得到模型保存时迭代的轮数 # global_step = ckpt.model_checkpoint_path.split('/')[-1].split('-')[-1] prediction,y_ = sess.run([prediction,y]) p = math.pow(e, y_[0][prediction]) / (math.pow(e, y_[0][0])+math.pow(e, y_[0][1])+math.pow(e, y_[0][2])+math.pow(e, y_[0][3])+math.pow(e, y_[0][4])+math.pow(e, y_[0][5]) +math.pow(e, y_[0][6])+math.pow(e, y_[0][7])+math.pow(e, y_[0][8])+math.pow(e, y_[0][9])+math.pow(e, y_[0][10])+math.pow(e, y_[0][11]) +math.pow(e, y_[0][12])+math.pow(e, y_[0][13])+math.pow(e, y_[0][14])+math.pow(e, y_[0][15])+math.pow(e, y_[0][16])+math.pow(e, y_[0][17]) +math.pow(e, y_[0][18])+math.pow(e, y_[0][19])+math.pow(e, y_[0][20])+math.pow(e, y_[0][21])+math.pow(e, y_[0][22])+math.pow(e, y_[0][23]) +math.pow(e, y_[0][24])+math.pow(e, y_[0][25])+math.pow(e, y_[0][26])+math.pow(e, y_[0][27])+math.pow(e, y_[0][28])+math.pow(e, y_[0][29]) +math.pow(e, y_[0][30])+math.pow(e, y_[0][31])+math.pow(e, y_[0][32])+math.pow(e, y_[0][33])+math.pow(e, y_[0][34])+math.pow(e, y_[0][35]) +math.pow(e, y_[0][36])+math.pow(e, y_[0][37])+math.pow(e, y_[0][38])+math.pow(e, y_[0][39])+math.pow(e, y_[0][40])+math.pow(e, y_[0][41]) +math.pow(e, y_[0][42])+math.pow(e, y_[0][43])+math.pow(e, y_[0][44])+math.pow(e, y_[0][45])+math.pow(e, y_[0][46])+math.pow(e, y_[0][47]) +math.pow(e, y_[0][48])+math.pow(e, y_[0][49])+math.pow(e, y_[0][50])+math.pow(e, y_[0][51])+math.pow(e, y_[0][52])+math.pow(e, y_[0][53]) +math.pow(e, y_[0][54])+math.pow(e, y_[0][55])+math.pow(e, y_[0][56])+math.pow(e, y_[0][57])+math.pow(e, y_[0][58])+math.pow(e, y_[0][59]) +math.pow(e, y_[0][60])+math.pow(e, y_[0][61])+math.pow(e, y_[0][62])+math.pow(e, y_[0][63])+math.pow(e, y_[0][64])) if prediction == 0: prediction = '0' elif prediction == 1: prediction = '1' elif prediction == 2: prediction = '2' elif prediction == 3: prediction = '3' elif prediction == 4: prediction = '4' elif prediction == 5: prediction = '5' elif prediction == 6: prediction = '6' elif prediction == 7: prediction = '7' elif prediction == 8: prediction = '8' elif prediction == 9: prediction = '9' elif prediction == 10: prediction = 'A' elif prediction == 11: prediction = 'B' elif prediction == 12: prediction = 'C' elif prediction == 13: prediction = 'D' elif prediction == 14: prediction = 'E' elif prediction == 15: prediction = 'F' elif prediction == 16: prediction = 'G' elif prediction == 17: prediction = 'H' elif prediction == 18: prediction = 'J' elif prediction == 19: prediction = 'K' elif prediction == 20: prediction = 'L' elif prediction == 21: prediction = 'M' elif prediction == 22: prediction = 'N' elif prediction == 23: prediction = 'P' elif prediction == 24: prediction = 'Q' elif prediction == 25: prediction = 'R' elif prediction == 26: prediction = 'S' elif prediction == 27: prediction = 'T' elif prediction == 28: prediction = 'U' elif prediction == 29: prediction = 'V' elif prediction == 30: prediction = 'W' elif prediction == 31: prediction = 'X' elif prediction == 32: prediction = 'Y' elif prediction == 33: prediction = 'Z' elif prediction == 34: prediction = '藏' elif prediction == 35: prediction = '川' elif prediction == 36: prediction = '鄂' elif prediction == 37: prediction = '甘' elif prediction == 38: prediction = '赣' elif prediction == 39: prediction = '桂' elif prediction == 40: prediction = '贵' elif prediction == 41: prediction = '黑' elif prediction == 42: prediction = '沪' elif prediction == 43: prediction = '吉' elif prediction == 44: prediction = '冀' elif prediction == 45: prediction = '津' elif prediction == 46: prediction = '晋' elif prediction == 47: prediction = '京' elif prediction == 48: prediction = '辽' elif prediction == 49: prediction = '鲁' elif prediction == 50: prediction = '蒙' elif prediction == 51: prediction = '闽' elif prediction == 52: prediction = '宁' elif prediction == 53: prediction = '青' elif prediction == 54: prediction = '琼' elif prediction == 55: prediction = '陕' elif p