谷歌以图搜图验证码识别.zip

# -*- coding: utf-8 -*- import argparse import json import operator import os import random from keras.callbacks import EarlyStopping from skimage import exposure import cv2 import imageio import matplotlib.pyplot as plt import numpy as np import pandas as pd import tensorflow as tf from keras import backend as K from keras.layers import Flatten, BatchNormalization from keras.layers import Input from keras.layers import concatenate from keras.layers.convolutional import AveragePooling2D from keras.layers.convolutional import Conv2D from keras.layers.convolutional import MaxPooling2D from keras.layers.core import Activation from keras.layers.core import Dense from keras.layers.core import Dropout from keras.models import Model from keras.preprocessing.image import ImageDataGenerator from keras.preprocessing.image import img_to_array from keras.utils.np_utils import to_categorical from sklearn.model_selection import train_test_split from sklearn.preprocessing import LabelEncoder gpus = tf.config.list_physical_devices(device_type='GPU') for gpu in gpus: tf.config.experimental.set_memory_growth(device=gpu, enable=True) src = "data\\" model_src = "model\\" photo_dir = r'photo/' if not os.path.exists(src): os.makedirs(src) if not os.path.exists(model_src): os.makedirs(model_src) if not os.path.exists(photo_dir): os.makedirs(photo_dir) # 参数配置 ap = argparse.ArgumentParser() # 输出文件夹路径 ap.add_argument("-m", "--model", type=str, default=model_src, help="path to output model") # 输出loss的迭代图片 ap.add_argument("-p", "--plot", type=str, default="plot.png", help="path to output loss/accuracy plot") args = vars(ap.parse_args()) def gasuss_noise(image, mean=0, var=0.001): ''' 手动添加高斯噪声 mean : 均值 var : 方差 ''' image = np.array(image / 255, dtype=float) noise = np.random.normal(mean, var ** 0.5, image.shape) # 正态分布 out = image + noise if out.min() < 0: low_clip = -1. else: low_clip = 0. out = np.clip(out, low_clip, 1.0) out = np.uint8(out * 255) return out def sp_noise(image, prob): ''' 手动添加椒盐噪声 prob:噪声比例 ''' output = np.zeros(image.shape, np.uint8) thres = 1 - prob for i in range(image.shape[0]): for j in range(image.shape[1]): rdn = random.random() if rdn < prob: output[i][j] = 0 elif rdn > thres: output[i][j] = 255 else: output[i][j] = image[i][j] return output def trans(img): # 转换颜色 img2 = img Size = img.shape width = Size[0] length = Size[1] list1_red = [0, 0, 128] list2_green = [0, 128, 0] for i in range(0, width): for j in range(0, length): if operator.eq(img[i][j].tolist(), list1_red) == True: img2[i][j] = np.array([0, 128, 0]) elif operator.eq(img[i][j].tolist(), list2_green) == True: img2[i][j] = np.array([0, 0, 128]) return img2 def readTxt(filepath): try: with open(filepath, 'r', encoding='gbk') as f: lines = [] for one in f: one = one.rstrip("\n\t") lines.append(one) f.close() return lines except: with open(filepath, 'r', encoding='utf-8') as f: lines = [] for one in f: one = one.rstrip("\n\t") lines.append(one) f.close() return lines # 写Json文件，写一行就换行，追加方式 def writeJson(relate_record, src): Json_str = json.dumps(relate_record, ensure_ascii=False) with open(src, 'w') as Json_file: Json_file.write(Json_str) Json_file.close() class MiniGoogLeNet: @staticmethod def conv_module(x, K, kX, kY, stride, chanDim, padding="same"): # x表示输入数据，K表示conv的filter的数量，KX,KY表示kernel_size # define a CONV => BN => RELU pattern,我们严格按照原论文的说法，使用CONV => BN => RELU的顺序，但是实际上，CONV => Relu => BN的效果会更好一些 x = Conv2D(K, (kX, kY), strides=stride, padding=padding)(x) x = BatchNormalization(axis=chanDim)(x) x = Activation("relu")(x) # return the block return x @staticmethod def inception_module(x, numK1_1, numK3_3, chanDim): # x表示输入数据,numK1_1,numK3_3表示kernel的filter的数量，chanDim：first_channel or last_channel conv1_1 = MiniGoogLeNet.conv_module(x, numK1_1, 1, 1, (1, 1), chanDim) conv3_3 = MiniGoogLeNet.conv_module(x, numK3_3, 3, 3, (1, 1), chanDim) x = concatenate([conv1_1, conv3_3], axis=chanDim) # 将conv1_1和conv3_3串联到一起 return x @staticmethod def downsample_module(x, K, chanDim): # K表示conv的filter的数量 conv3_3 = MiniGoogLeNet.conv_module(x, K, 3, 3, (2, 2), chanDim, padding='valid') # padding=same表示：出输入和输出的size是相同的，由于加入了padding，如果是padding=valid，那么padding=0 pool = MaxPooling2D((3, 3), strides=(2, 2))(x) x = concatenate([conv3_3, pool], axis=chanDim) # 将conv3_3和maxPooling串到一起 return x @staticmethod def build(width, height, depth, classes): inputShape = (height, width, depth) # keras默认channel last，tf作为backend chanDim = -1 # if we are using "channels first", update the input shape # and channels dimension if K.image_data_format() == "channels_first": inputShape = (depth, height, width) chanDim = 1 # define the model input and first CONV module inputs = Input(shape=inputShape) x = MiniGoogLeNet.conv_module(inputs, 96, 3, 3, (1, 1), chanDim) # two Inception modules followed by a downsample module x = MiniGoogLeNet.inception_module(x, 32, 32, chanDim) # 第一个分叉 x = MiniGoogLeNet.inception_module(x, 32, 48, chanDim) # 第二个分叉 x = MiniGoogLeNet.downsample_module(x, 80, chanDim) # 第三个分叉，含有maxpooling # four Inception modules followed by a downsample module x = MiniGoogLeNet.inception_module(x, 112, 48, chanDim) x = MiniGoogLeNet.inception_module(x, 96, 64, chanDim) x = MiniGoogLeNet.inception_module(x, 80, 80, chanDim) x = MiniGoogLeNet.inception_module(x, 48, 96, chanDim) x = MiniGoogLeNet.downsample_module(x, 96, chanDim) # two Inception modules followed by global POOL and dropout x = MiniGoogLeNet.inception_module(x, 176, 160, chanDim) x = MiniGoogLeNet.inception_module(x, 176, 160, chanDim) # 输出是（7×7×（160+176）） x = AveragePooling2D((7, 7))(x) # 经过平均池化之后变成了（1*1*376） x = Dropout(0.5)(x) # softmax classifier x = Flatten()(x) # 特征扁平化 x = Dense(classes)(x) # 全连接层，进行多分类,形成最终的10分类 x = Activation("softmax")(x) # create the model model = Model(inputs, x, name="googlenet") # return the constructed network architecture return model def build_train(path): files = os.listdir(path) count = len(files) for imagePath in files: try: image = imageio.imread(path + '/' + imagePath) except: continue image = cv2.resize(image, (image_shape, image_shape)) # 加噪音 out3 = gasuss_noise(image) out4 = sp_noise(image,