Python手写数字识别
跨考上浙大 人气:01.效果图
有点low，轻喷
点击选择图片会优先从当前目录查找
2.数据集
这部分我是对MNIST数据集进行处理保存
对应代码：
import tensorflow as tf
import matplotlib.pyplot as plt
import cv2
from PIL import Image
import numpy as np
from scipy import misc
(x_train_all,y_train_all),(x_test,y_test) = tf.keras.datasets.mnist.load_data()
x_valid,x_train = x_train_all[:5000],x_train_all[5000:]
y_valid,y_train = y_train_all[:5000],y_train_all[5000:]
print(x_valid.shape,y_valid.shape)
print(x_train.shape,y_train.shape)
print(x_test.shape,y_test.shape)
#读取单张图片
def show_single_img(img_arr,len=100,path='/Users/zhangcaihui/Desktop/case/jpg/'):
for i in range(len):#我这种写法会进行覆盖，只能保存10张照片，想保存更多的数据自己看着改
new_im = Image.fromarray(img_arr[i]) # 调用Image库，数组归一化
#new_im.show()
#plt.imshow(img_arr) # 显示新图片
label=y_train[i]
new_im.save(path+str(label)+'.jpg') # 保存图片到本地
#显示多张图片
def show_imgs(n_rows,n_cols,x_data,y_data):
assert len(x_data) == len(y_data)
assert n_rows * n_cols < len(x_data)
plt.figure(figsize=(n_cols*1.4,n_rows*1.6))
for row in range(n_rows):
for col in range(n_cols):
index = n_cols * row + col
plt.subplot(n_rows,n_cols,index+1)
plt.imshow(x_data[index],cmap="binary",interpolation="nearest")
plt.axis("off")
plt.show()
#show_imgs(2,2,x_train,y_train)
show_single_img(x_train)
3.关于模型
我保存了了之前训练好的模型，用来加载预测
关于tensorflow下训练神经网络模型：手把手教你，MNIST手写数字识别
训练好的模型model.save(path)即可
4.关于GUI设计
1）排版
#ui_openimage.py
# -*- coding: utf-8 -*-
# from PyQt5 import QtCore, QtGui, QtWidgets
# from PyQt5.QtCore import Qt
import sys,time
from PyQt5 import QtGui, QtCore, QtWidgets
from PyQt5.QtWidgets import *
from PyQt5.QtCore import *
from PyQt5.QtGui import *
class Ui_Form(object):
def setupUi(self, Form):
Form.setObjectName("Form")
Form.resize(1144, 750)
self.label_1 = QtWidgets.QLabel(Form)
self.label_1.setGeometry(QtCore.QRect(170, 130, 351, 251))
self.label_1.setObjectName("label_1")
self.label_2 = QtWidgets.QLabel(Form)
self.label_2.setGeometry(QtCore.QRect(680, 140, 351, 251))
self.label_2.setObjectName("label_2")
self.btn_image = QtWidgets.QPushButton(Form)
self.btn_image.setGeometry(QtCore.QRect(270, 560, 93, 28))
self.btn_image.setObjectName("btn_image")
self.btn_recognition = QtWidgets.QPushButton(Form)
self.btn_recognition.setGeometry(QtCore.QRect(680,560,93,28))
self.btn_recognition.setObjectName("bnt_recognition")
#显示时间按钮
self.bnt_timeshow = QtWidgets.QPushButton(Form)
self.bnt_timeshow.setGeometry(QtCore.QRect(900,0,200,50))
self.bnt_timeshow.setObjectName("bnt_timeshow")
self.retranslateUi(Form)
self.btn_image.clicked.connect(self.slot_open_image)
self.btn_recognition.clicked.connect(self.slot_output_digital)
self.bnt_timeshow.clicked.connect(self.buttonClicked)
self.center()
QtCore.QMetaObject.connectSlotsByName(Form)
def retranslateUi(self, Form): #设置文本填充label、button
_translate = QtCore.QCoreApplication.translate
Form.setWindowTitle(_translate("Form", "数字识别系统"))
self.label_1.setText(_translate("Form", "点击下方按钮"))
self.label_1.setStyleSheet('font:50px;')
self.label_2.setText(_translate("Form", "0~9"))
self.label_2.setStyleSheet('font:50px;')
self.btn_image.setText(_translate("Form", "选择图片"))
self.btn_recognition.setText(_translate("From","识别结果"))
self.bnt_timeshow.setText(_translate("Form","当前时间"))
# 状态条显示时间模块
def buttonClicked(self): # 动态显示时间
timer = QTimer(self)
timer.timeout.connect(self.showtime)
timer.start()
def showtime(self):
datetime = QDateTime.currentDateTime()
time_now = time.strftime('%Y-%m-%d %H:%M:%S', time.localtime())
#self.statusBar().showMessage(time_now)
#self.bnt_timeshow.setFont(QtGui.QFont().setPointSize(100))
self.bnt_timeshow.setText(time_now)
def center(self):#窗口放置中央
screen = QDesktopWidget().screenGeometry()
size = self.geometry()
self.move((screen.width() - size.width()) / 2,
(screen.height() - size.height()) / 2)
def keyPressEvent(self, e):
if e.key() == Qt.Key_Escape:
self.close()
2）直接运行这个文件（调用1）
#ui_main.py
import random
from PyQt5.QtWidgets import QFileDialog
from PyQt5.QtGui import QPixmap
from ui_openimage import Ui_Form
import sys
from PyQt5 import QtWidgets, QtGui
from PyQt5.QtWidgets import QMainWindow, QTextEdit, QAction, QApplication
import os,sys
from PyQt5.QtCore import Qt
import tensorflow
from tensorflow.keras.models import load_model
from tensorflow.keras.datasets import mnist
from tensorflow.keras import models
from tensorflow.keras import layers
from tensorflow.keras.utils import to_categorical
import tensorflow.keras.preprocessing.image as image
import matplotlib.pyplot as plt
import numpy as np
import cv2
import warnings
warnings.filterwarnings("ignore")
class window(QtWidgets.QMainWindow,Ui_Form):
def __init__(self):
super(window, self).__init__()
self.cwd = os.getcwd()
self.setupUi(self)
self.labels = self.label_1
self.img=None
def slot_open_image(self):
file, filetype = QFileDialog.getOpenFileName(self, '打开多个图片', self.cwd, "*.jpg, *.png, *.JPG, *.JPEG, All Files(*)")
jpg = QtGui.QPixmap(file).scaled(self.labels.width(), self.labels.height())
self.labels.setPixmap(jpg)
self.img=file
def slot_output_digital(self):
'''path为之前保存的模型路径'''
path='/Users/zhangcaihui/PycharmProjects/py38_tf/DL_book_keras/save_the_model.h5'
model= load_model(path)
#防止不上传数字照片而直接点击识别
if self.img==None:
self.label_2.setText('请上传照片！')
return
img = image.load_img(self.img, target_size=(28, 28))
img = img.convert('L')#转灰度图像
x = image.img_to_array(img)
#x = abs(255 - x)
x = np.expand_dims(x, axis=0)
print(x.shape)
x = x / 255.0
prediction = model.predict(x)
print(prediction)
output = np.argmax(prediction, axis=1)
print("手写数字识别为：" + str(output[0]))
self.label_2.setText(str(output[0]))
if __name__ == "__main__":
app = QtWidgets.QApplication(sys.argv)
my = window()
my.show()
sys.exit(app.exec_())
5.缺点
界面low
只能识别单个数字
其实可以将多数字图片进行裁剪分割，这就涉及到制作数据集了
6.遗留问题
我自己手写的数据照片处理成28281送入网络预测，识别结果紊乱。
反思：自己写的数据是RGB，且一张几KB，图片预处理后，按28*28读入失真太严重了，谁有好的方法可以联系我！！！
其他的水果识别系统，手势识别系统啊，改改直接套！
加载全部内容