這個模型是經過簡單訓練，只能識別四個英文字母：ABCD，是我為了提取答案才訓練的，因為科大，百度，以及騰訊的識別不是很準，所以自己訓練自己使用，希望能夠給你帶來方便，下面是訓練好的model,如多你感覺圖片的型別相似，可直接使用，
1、圖片形式，盡量png
2、模型
如果需要可以加我qq：1851999452
3、
1 import os 2 import numpy as np 4 import cv2 5 import matplotlib.pyplot as plt 6 import tensorflow as tf 7 import keras 8 from keras.models import Model,load_model 9 from keras.layers import Conv2D,Dense,Activation,Input,MaxPooling2D,Flatten 10 from sklearn.model_selection import train_test_split 11 # file_dir = '/home/youxue/BorD' 12 # B = [] 13 # A =[] 14 # C =[] 15 # label_A=[] 16 # label_C=[] 17 # label_B = [] 18 # D = [] 19 # label_D = [] 20 # for file in os.listdir(file_dir): 21 # if file == 'A': 22 # for img in os.listdir(os.path.join(file_dir, file)): 23 # A.append(file_dir + '/' + file + '/' + img) 24 # label_A.append(0) 25 # elif file == 'B': 26 # for img in os.listdir(os.path.join(file_dir, file)): 27 # B.append(file_dir + '/' + file + '/' + img) 28 # label_B.append(1) 29 # elif file == 'C': 30 # for img in os.listdir(os.path.join(file_dir, file)): 31 # C.append(file_dir + '/' + file + '/' + img) 32 # label_C.append(2) 33 # else: 34 # for img in os.listdir(os.path.join(file_dir, file)): 35 # D.append(file_dir + '/' + file + '/' + img) 36 # label_D.append(3) 37 # 38 # print('There are %d B\n There are %d D' % (len(B), len(D))) 39 # 40 # image_list = np.hstack((A,B,C,D)) 41 # label_list = np.hstack((label_A,label_B,label_C, label_D)) 42 # temp = np.array([image_list, label_list]) 43 # temp = temp.transpose() 44 # 45 # 46 # np.random.shuffle(temp) 47 # png_list=[] 48 # image_list = list(temp[:, 0]) 49 # label_list = list(temp[:, 1]) 50 # label_list = [int(i) for i in label_list] 51 # for png in image_list: 52 # png_list.append(cv2.imread(png)) 53 # print(png_list[100].shape) 54 # png = np.array(png_list) 55 # label = np.array(label_list) 56 # x_train,x_test,y_train,y_test=train_test_split(png,label,test_size=0.1) 57 # 58 # # plt.imshow(x_train[11]) 59 # # plt.show() 60 # 61 # x_train = x_train/255. 62 # x_test = x_test/255. 63 # 64 # y_train = np.eye(4)[y_train.reshape(-1)] 65 # y_test = np.eye(4)[y_test.reshape(-1)] 66 # print(x_train.shape,y_train.shape) 67 68 # input_shape= (24,810,3) 69 # input_data = https://www.cnblogs.com/beisi/p/Input(input_shape) 70 # layer1 = Conv2D(32,(3,3),padding='same',activation='relu')(input_data) 71 # layer2 = MaxPooling2D(pool_size=(2,2))(layer1) 72 # layer3 = Conv2D(32,(3,3),padding='same',activation='relu')(layer2) 73 # layer4 = MaxPooling2D(pool_size=(2,2))(layer3) 74 # layer5 = Conv2D(32,(3,3),padding='same',activation='relu')(layer4) 75 # layer6 = Flatten()(layer5) 76 # layer7 = Dense(4)(layer6) 77 # pred= Activation('softmax')(layer7) 78 # model= Model(input=input_data,output=pred) 79 # model.summary() 80 # model.compile(optimizer='adam',loss='categorical_crossentropy',metrics=['accuracy']) 81 # model.fit(x_train,y_train,batch_size=32,epochs=20,validation_data=https://www.cnblogs.com/beisi/p/(x_test,y_test)) 82 # model.save('ABCD.model') 83 model=load_model('ABCD.model') 84 #這里填寫自己的存放路徑 85 with open('id.txt','a+') as f: 86 #這里填寫你需要識別圖片的路徑 87 path='' 88 for png in os.listdir(path): 89 pn= cv2.imread(os.path.join(path,png)) 90 ids_1 = png.split('answer')[0] 91 print(type(ids_1)) 92 #png2= cv2.imread('/home/youxue/4907988answer.png') 93 # plt.imshow(pn) 94 # plt.show() 95 img=[] 96 img.append(pn) 97 #img.append(png2) 98 img=np.array(img) 99 img = img/255.100 101 # print(img.shape)102 pre=model.predict(img)103 for c in pre:104 temp= max(c[0],c[1],c[2],c[3])105 if temp == c[0]:106 ABCD = 'A'107 elif temp == c[1]:108 ABCD = 'B'109 elif temp == c[2]:110 ABCD = 'C'111 else:112 ABCD = 'D'113 114 f.write(ids_1 + '--' + ABCD + '\n')
轉載請註明出處，本文鏈接：https://www.uj5u.com/caozuo/27230.html
標籤：Windows
上一篇：解決windows 1903 alt + tab 切換卡頓
下一篇：mybatis入門百分百