kivy手写数字识别app

使用python kivy+opencv实现手写数字识别的安卓app。
app下载：
https://pan.baidu.com/s/1l1iqMYmZeHuq-c_mg15F0Q
提取码：lvev

如有问题，欢迎下方评论！

kivy安卓app实现数字识别

代码部分：

from kivy.app import App
from kivy.uix.boxlayout import BoxLayout
from kivy.graphics.texture import Texture
from kivy.uix.camera import Camera
from kivy.uix.image import Image
from kivy.uix.button import Button
from kivy.clock import Clock
from kivy.graphics import Color,Rectangle
import numpy as np
import cv2
from android.permissions import request_permissions, Permissionrequest_permissions([Permission.CAMERA,Permission.WRITE_EXTERNAL_STORAGE,Permission.READ_EXTERNAL_STORAGE
])net = cv2.dnn.readNetFromTensorflow('model.pb')
def detect(src,net):try:src_copy=src.copy()gray=cv2.cvtColor(src,cv2.COLOR_RGB2GRAY)# thred=np.where(gray>140,255,0).astype('uint8')_,thred=cv2.threshold(gray,100,255,cv2.THRESH_BINARY_INV)# thred=255-thred#开闭运算k = np.ones((5, 5), np.uint8)#thred=cv.dilate(thred,k)thred = cv2.morphologyEx(thred, cv2.MORPH_CLOSE, k)thred=cv2.dilate(thred,k)# cv2.imshow('a123', thred)# cv2.waitKey(0)cnts=cv2.findContours(thred,cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_SIMPLE)[0]# print('轮廓数=',len(cnts))if len(cnts)>100:return src_copy,thrednam=[]for i in cnts:area = cv2.contourArea(i)if area<50:continue# if aa!=1:#     b=20# else:b=15# print('面积=',area)b=50if area>=b:x, y, w, h = cv2.boundingRect(i)x2=x-10y2=y-10w2=w+20h2=h+20cv2.rectangle(src_copy,(x2,y2),(x2+w2,y2+h2),(0,255,0),2)lkuo=thred[y:y+h,x:x+w]da = max(h, w)rate = da / 40ro = cv2.resize(lkuo, (int(w / rate), int(h / rate)))h, w = ro.shapet, b = int((43 - h) / 2), 43 - h - int((43 - h) / 2)l, r = int((43 - w) / 2), 43 - w - int((43 - w) / 2)ro = cv2.copyMakeBorder(ro, t, b, l, r, cv2.BORDER_CONSTANT, value=0)ro = cv2.resize(ro, (40, 40))ro = np.where(ro > 0, 255, 0).astype('float32')# cv.imshow('a123', ro)# cv.waitKey(0)ro = ro / 255# ro=np.reshape(ro,(1,40,40,1))blob=ro.reshape((1,1,40,40))net.setInput(blob)out=net.forward()#图上标出数字number=np.argmax(out,axis=-1)nam.append((number[0],x2, y2-5,h/40))#cv.putText(src_copy, str(number[0]), (x2, y2-5), cv.FONT_HERSHEY_SIMPLEX, h/40, (0, 0, 255), 2)#print('opencv预测结果：',np.argmax(out,axis=-1),'tensorflow预测结果：',np.argmax(out2,axis=-1))if len(nam)==0:return src_copy,thredfor ii in nam:cv2.putText(src_copy, str(ii[0]), (ii[1], ii[2]), cv2.FONT_HERSHEY_SIMPLEX, ii[3], (0, 0, 255), 2)return src_copy,thredexcept:return src,srcclass MyLayout(BoxLayout):def __init__(self):super(MyLayout, self).__init__()self.orientation='vertical'self.image1 = Image()self.image1.color=1,1,1,0self.image1.allow_stretch=Trueself.btn=Button(text='OPEN CAMERA')with self.canvas:Color(1,1,1,1)self.rect=Rectangle(pos=self.pos,size=self.size,source='11.jpg')self.bind(pos=self.update,size=self.update)self.padding=[0,0,0,10]self.spacing=10self.btn.size_hint=.3,.06self.btn.pos_hint={'center_x':0.5,'center_y':.5}self.btn.background_color=1,1,1,.8# self.btn.size_hint_y=.1# self.btn.size_hint_x=.2self.btn.bind(on_press=self.press_btn)self.camera = Camera(resolution=(640, 480))self.camera.play=Falseself.add_widget(self.image1)self.add_widget(self.btn)Clock.schedule_interval(self.pr, 0)def update(self,*args):self.rect.pos=self.posself.rect.size=self.sizedef press_btn(self,arg):self.camera.play=not self.camera.playif self.camera.play:self.image1.color=1,1,1,1self.btn.text='CLOSE CAMERA'else:self.image1.color = 1, 1, 1, 0self.btn.text = 'OPEN CAMERA'def pr(self, nap):if self.camera.texture==None:returnimg_rgb=self.texture_to_numpy(self.camera.texture)[:,:,:3]img_rgb_90=np.rot90(img_rgb,-1)img_rgb_zy=img_rgb_90[:,::-1,:]out,_=detect(img_rgb_zy,net)#cv2.putText(img_rgba_copy, str(self.counter), (100, 100), cv2.FONT_HERSHEY_SIMPLEX, 2, (0, 255, 0), 3)#texture1=self.numpy_to_texture(out)self.image1.texture=texture1# 将Texture格式的图片对象转换为numpydef texture_to_numpy(self, data):image = np.asarray(bytearray(data.pixels), dtype='uint8').reshape((data.height, data.width, 4))# r_chanel = np.copy(image[:, :, 0])# g_chanel = np.copy(image[:, :, 1])# b_chanel = np.copy(image[:, :, 2])# image[:, :, 0] = b_chanel# image[:, :, 1] = g_chanel# image[:, :, 2] = r_chanel#返回rgba图像return image# 将numpy格式图片对象转为Texturedef numpy_to_texture(self, frame):# frame=cv2.imread("ddd.jpg")buf1 = cv2.flip(frame, 0)buf = buf1.tostring()image_texture = Texture.create(size=(frame.shape[1], frame.shape[0]), colorfmt='rgb')image_texture.blit_buffer(buf, colorfmt='rgb', bufferfmt='ubyte')return image_textureclass MyApp(App):def build(self):return MyLayout()if __name__ == '__main__':MyApp().run()

kivy手写数字识别app相关推荐

python手写多个字母识别_一个带界面的CNN手写数字识别，使用Python（tensorflow, kivy）实现...
CNN_Handwritten_Digit_Recognizer (CNN手写数字识别) A CNN handwritten digit recognizer with graphical UI, i ...
学习笔记CB009:人工神经网络模型、手写数字识别、多层卷积网络、词向量、word2vec...
人工神经网络,借鉴生物神经网络工作原理数学模型. 由n个输入特征得出与输入特征几乎相同的n个结果,训练隐藏层得到意想不到信息.信息检索领域,模型训练合理排序模型,输入特征,文档质量.文档点击历史.文档 ...
将tensorflow训练好的模型移植到Android (MNIST手写数字识别)
将tensorflow训练好的模型移植到Android (MNIST手写数字识别) [尊重原创,转载请注明出处]https://blog.csdn.net/guyuealian/article/det ...
持久化的基于L2正则化和平均滑动模型的MNIST手写数字识别模型
持久化的基于L2正则化和平均滑动模型的MNIST手写数字识别模型觉得有用的话,欢迎一起讨论相互学习~Follow Me 参考文献Tensorflow实战Google深度学习框架实验平台: Tens ...
Tensorflow 神经网络作业手写数字识别训练、回测准确率
大白话讲解卷积神经网络工作原理,推荐一个bilibili的讲卷积神经网络的视频,up主从youtube搬运过来,用中文讲了一遍. 这篇文章是 TensorFlow 2.0 Tutorial 入门教程的 ...
TensorFlow 2.0 mnist手写数字识别(CNN卷积神经网络)
TensorFlow 2.0 (五) - mnist手写数字识别(CNN卷积神经网络) 源代码/数据集已上传到 Github - tensorflow-tutorial-samples 大白话讲解卷积 ...
机器学习框架ML.NET学习笔记【5】多元分类之手写数字识别（续）
一.概述上一篇文章我们利用ML.NET的多元分类算法实现了一个手写数字识别的例子,这个例子存在一个问题,就是输入的数据是预处理过的,很不直观,这次我们要直接通过图片来进行学习和判断.思路很简单,就是 ...
手写数字识别的小优化
在用KNN实现手写数字识别的时候突发奇想是否可以根据数字的特点对其进行一个分类,以此来提高判断的准确率.经过许多天的改进与完善终于实现了此算法,就称其为洞拐法吧.(第一次写,有许多不足之处,还望多多包 ...
单层神经网络实现手写数字识别
Mnist手写数字识别前言 Mnist数据集可以从官网下载,网址: http://yann.lecun.com/exdb/mnist/ 下载下来的数据集被分成两部分: 55000行的训练数据集(mn ...

kivy手写数字识别app

kivy手写数字识别app相关推荐

最新文章

热门文章