Python image classify code

import tensorflow as tf from tensorflow import keras import cv2 import numpy as np import glob import os import sys import imutils import random import pymssql import pyodbc from PIL import Image, ImageTk from tensorflow.keras import losses,layers,optimizers from tensorflow.keras.callbacks import EarlyStopping from PyQt5.QtWidgets import QApplication, QMainWindow, QDockWidget, QTextEdit, QLabel, QWidget, QPushButton, QVBoxLayout, QListWidget, QLineEdit, QStyleFactory from PyQt5.QtCore import Qt, QTimer from PyQt5 import QtCore, QtGui, QtWidgets, uic from PyQt5.QtGui import QImage, QPixmap from PyQt5.QtSql import QSqlDatabase , QSqlQuery from pathlib import Path, PureWindowsPath QApplication.setStyle(QStyleFactory.create('Fusion')) #定义目标文件夹为全局变量，给各个函数进行调用 targetfilePath = '' isPredicting = True temp = '' count = 0 model = object trainingfilePath = '' mynet = object Camera_status = False saveImage = object #---------------------------------------------------------------------------------训练模块---------------------------------------------------------------------------------------------------# #训练参数设置 tf.random.set_seed(2222) np.random.seed(2222) os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2' assert tf.__version__.startswith('2.') #预处理训练图片 def preprocessTraining(x,y): x = tf.image.resize(x,[244,244]) x = tf.image.random_flip_left_right(x) x = tf.image.random_crop(x,[224,224,3]) x = tf.cast(x, dtype=tf.float32) / 255. x = normalize(x) y = tf.convert_to_tensor(y) y = tf.one_hot(y, depth=5) return x,y #加载训练图片以及Resize转换 def Data_Generation(filePath): X_data = [] Y_data = [] path_data = [] path_label = [] files = os.listdir(filePath) #循环加载文件 for file in files: if os.path.exists(trainingfilePath + '\\' + file+ '\\') == True: for path in glob.glob(trainingfilePath + '\\' + file + '\*.*'): if 'jpg' or 'png' or 'jpeg' in path: path_data.append(path) else: for path in glob.glob(trainingfilePath + '\*.*'): if 'jpg' or 'png' or 'jpeg' in path: path_data.append(path) #打乱数据 random.shuffle(path_data) #按照图片分类进行label标记 for paths in path_data: #print(paths) if 'Normal' in paths: path_label.append(0) elif 'BKG' in paths: path_label.append(1) img = cv2.imdecode(np.fromfile(paths, dtype=np.uint8), -1) #img = cv2.imread(paths) img = cv2.cvtColor(img,cv2.COLOR_BGR2RGB) img = cv2.resize(img,(224,224)) X_data.append(img) L = len(path_data) Y_data = path_label X_data = np.array(X_data,dtype=float) Y_data = np.array(Y_data,dtype='uint8') X_train = X_data[0:int(L*0.6)] Y_train = Y_data[0:int(L*0.6)] X_valid = X_data[int(L*0.6):int(L*0.8)] Y_valid = Y_data[int(L*0.6):int(L*0.8)] X_test = X_data[int(L*0.8):] Y_test = Y_data[int(L*0.8):] return X_train,Y_train,X_valid,Y_valid,X_test,Y_test,L #获取训练集，测试集 def trainingModel(self): #文件加载路径 trainingfilePath = targetfilePath if len(trainingfilePath) == 0: self.logText.setText('请先选择文件夹再开始训练!') return else: self.logText.setText('开始训练!') X_train,Y_train,X_valid,Y_valid,X_test,Y_test,L = Data_Generation(trainingfilePath) batchsz = 32 #print(shape(X_data), shape(Y_data)) train_db = tf.data.Dataset.from_tensor_slices((X_train,Y_train)) train_db = train_db.shuffle(10000).map(preprocessTraining).batch(batchsz) valid_db = tf.data.Dataset.from_tensor_slices((X_valid,Y_valid)) valid_db = valid_db.map(preprocessTraining).batch(batchsz) test_db = tf.data.Dataset.from_tensor_slices((X_test,Y_test)) test_db = test_db.map(preprocessTraining).batch(batchsz) #这里使用了自带的DenseNet121网络 你也可以用keras.Sequential DIY模型 net = keras.applications.DenseNet121(weights='imagenet',include_top=False,pooling='max') net.trainable = False global mynet mynet = keras.Sequential([ net, layers.Dense(1024,activation='relu'), layers.BatchNormalization(), #BN层 标准化数据 layers.Dropout(rate=0.2), layers.Dense(5)]) mynet.build(input_shape=(4,224,224,3)) mynet.summary() #防止过拟合 early_stopping = EarlyStopping(monitor='val_accuracy', min_delta=0.01, patience=3) mynet.compile(optimizer=optimizers.Adam(lr=1e-3), loss=losses.CategoricalCrossentropy(from_logits=True), metrics=['accuracy']) history = mynet.fit(train_db, validation_data=valid_db, validation_freq=1, epochs=50, callbacks=[early_stopping]) history = history.history mynet.evaluate(test_db) self.logText.setText('训练完成!') #保存模型 def saveModel(self): FolderName = QtWidgets.QFileDialog.getExistingDirectory() if '/' in FolderName : # 用\替换/，注意'\\'的用法， FolderName.replace('/', '\\') #训练结束以后保存mmodel文件到本地方便做图片分类的时候直接调用 #保存model成 .h5格式 里面包含了模型结构和训练好的模型参数 if len(FolderName) == 0: self.logText.setText('未保存模型文件!') else: mynet.save(FolderName + '\densenet.h5') self.logText.setText('模型文件已保存!') #---------------------------------------------------------------------------------识别模块---------------------------------------------------------------------------------------------------# #图片归一化 def normalize(x): img_mean = tf.constant([0.485, 0.456, 0.406]) img_std = tf.constant([0.229, 0.224, 0.225]) x = (x - img_mean)/img_std return x #初始化图片 def preprocess(x): x = tf.expand_dims(x,axis=0) x = tf.cast(x, dtype=tf.float32) / 255. x = normalize(x) return x #加载模型 def loadModel(self): File = QtWidgets.QFileDialog.getOpenFileName() modelFile = File[0] if '/' in modelFile: # 用\替换/，注意'\\'的用法， modelFile.replace('/', '\\') if len(modelFile) == 0: self.logText.setText('未加载模型文件!') else: self.modelSelectText.setText(modelFile) network = keras.models.load_model(modelFile) network.summary() self.logText.setText('检测模型加载成功：' + modelFile) global model model = network #打开目标文件夹 def openFileFolder(self): FolderName = QtWidgets.QFileDialog.getExistingDirectory() if '/' in FolderName : # 用\替换/，注意'\\'的用法， FolderName.replace('/', '\\') if len(FolderName) == 0 : self.logText.setText('未选择文件夹路径!') else : #print('目标文件夹:',FolderName) self.fileSelectText.setText(FolderName) self.logText.setText('目标文件夹:' + FolderName) #全局定义目标文件夹路径，更新路径 global targetfilePath targetfilePath = FolderName #图片识别主函数入口 def prediction(self, file, model): global temp global count #\\mysw1pda2002\PDASImages7\101\2023_06\MA836103\MRS1\0981_MRS1_1_MA836103_0161_102_0362_0007_1.PNG if model == object: self.logText.setText('未加载模型!') return label = ['Normal','BKG'] image = cv2.imdecode(np.fromfile(file, dtype=np.uint8), -1) img = image.copy() img = cv2.resize(img,(224,224)) img = cv2.cvtCo