CV2识别图片A,B,C,D

这个模型是经过简单训练，只能识别四个英文字母：ABCD，是我为了提取答案才训练的，因为科大，百度，以及腾讯的识别不是很准，所以自己训练自己使用，希望能够给你带来方便。下面是训练好的model,如多你感觉图片的类型相似，可直接使用。

1、图片形式，尽量png

2、模型

如果需要可以加我qq：1851999452

3、

import os
import numpy as np
import cv2
import matplotlib.pyplot as plt
import tensorflow as tf
import keras
from keras.models import Model,load_model
from keras.layers import Conv2D,Dense,Activation,Input,MaxPooling2D,Flatten
from sklearn.model_selection import train_test_split
# file_dir = '/home/youxue/BorD'
# B = []
# A =[]
# C =[]
# label_A=[]
# label_C=[]
# label_B = []
# D = []
# label_D = []
# for file in os.listdir(file_dir):
#     if file == 'A':
#         for img in os.listdir(os.path.join(file_dir, file)):
#             A.append(file_dir + '/' + file + '/' + img)
#             label_A.append(0)
#     elif file == 'B':
#         for img in os.listdir(os.path.join(file_dir, file)):
#             B.append(file_dir + '/' + file + '/' + img)
#             label_B.append(1)
#     elif file == 'C':
#         for img in os.listdir(os.path.join(file_dir, file)):
#             C.append(file_dir + '/' + file + '/' + img)
#             label_C.append(2)
#     else:
#         for img in os.listdir(os.path.join(file_dir, file)):
#             D.append(file_dir + '/' + file + '/' + img)
#             label_D.append(3)
#
# print('There are %d B
 There are %d D' % (len(B), len(D)))
#
# image_list = np.hstack((A,B,C,D))
# label_list = np.hstack((label_A,label_B,label_C, label_D))
# temp = np.array([image_list, label_list])
# temp = temp.transpose()
#
#
# np.random.shuffle(temp)
# png_list=[]
# image_list = list(temp[:, 0])
# label_list = list(temp[:, 1])
# label_list = [int(i) for i in label_list]
# for png in image_list:
#     png_list.append(cv2.imread(png))
# print(png_list[100].shape)
# png = np.array(png_list)
# label = np.array(label_list)
# x_train,x_test,y_train,y_test=train_test_split(png,label,test_size=0.1)
#
# # plt.imshow(x_train[11])
# # plt.show()
#
# x_train = x_train/255.
# x_test = x_test/255.
#
# y_train = np.eye(4)[y_train.reshape(-1)]
# y_test = np.eye(4)[y_test.reshape(-1)]
# print(x_train.shape,y_train.shape)

# input_shape= (24,810,3)
# input_data = Input(input_shape)
# layer1 = Conv2D(32,(3,3),padding='same',activation='relu')(input_data)
# layer2 = MaxPooling2D(pool_size=(2,2))(layer1)
# layer3 = Conv2D(32,(3,3),padding='same',activation='relu')(layer2)
# layer4 = MaxPooling2D(pool_size=(2,2))(layer3)
# layer5 = Conv2D(32,(3,3),padding='same',activation='relu')(layer4)
# layer6 = Flatten()(layer5)
# layer7 = Dense(4)(layer6)
# pred= Activation('softmax')(layer7)
# model= Model(input=input_data,output=pred)
# model.summary()
# model.compile(optimizer='adam',loss='categorical_crossentropy',metrics=['accuracy'])
# model.fit(x_train,y_train,batch_size=32,epochs=20,validation_data=(x_test,y_test))
# model.save('ABCD.model')
model=load_model('ABCD.model')
#这里填写自己的存放路径
with open('id.txt','a+') as f:
    #这里填写你需要识别图片的路径
    path=''
    for png in os.listdir(path):
        pn= cv2.imread(os.path.join(path,png))
        ids_1 = png.split('answer')[0]
        print(type(ids_1))
    #png2= cv2.imread('/home/youxue/4907988answer.png')
        # plt.imshow(pn)
        # plt.show()
        img=[]
        img.append(pn)
    #img.append(png2)
        img=np.array(img)
        img = img/255.

    # print(img.shape)
        pre=model.predict(img)
        for c in pre:
            temp= max(c[0],c[1],c[2],c[3])
            if temp == c[0]:
                ABCD = 'A'
            elif temp == c[1]:
                ABCD = 'B'
            elif temp == c[2]:
                ABCD = 'C'
            else:
                ABCD = 'D'
       
        f.write(ids_1 + '--' + ABCD + '
')