问题描述
import sys
from os import listdir,sep
import numpy as np
import pickle
from PIL import Image
import cv2
from sklearn.ensemble import RandomForestClassifier
import numpy as np
from PIL import Image
DEFAULT_IMAGE_SIZE = (45,45)
SPLIT_POINT_COEFF = 0.8
class PredictionService:
def __init__(self):
#self.dataset = datasetDIR
#self.testimage = testimage
pass
def getimageVector(self,image):
try:
#NOTE:from docs
#When translating a color image to black and white (mode “L”),the library uses the ITU-R 601-2 luma transform:
#L = R * 299/1000 + G * 587/1000 + B * 114/1000
imageGrayscale = Image.open(image).convert('L')
#resize image to default image size - 45 x 45
imageGrayscale = imageGrayscale.resize(DEFAULT_IMAGE_SIZE,Image.ANTIALIAS)
#
imageNP = np.array(imageGrayscale)
imgList = []
for line in imageNP:
for value in line:
imgList.append(value)
#imgList is 2025 long vector
return imgList
except Exception as e:
print("Error : {}".format(e))
return None
def addImagesToSet(self,rootPath,imageList,label,completeImageList = [],labelList = []):
dashes = ['-','/','-','\\']
counter = 0
for image in imageList:
print('[{}] Images loading...'.format(dashes[counter]))
counter = (counter + 1) % len(dashes)
completeImageList.append(self.getimageVector(rootPath + image))
labelList.append(label)
def getTrainingAndTestData(self,directoryPath):
dirList = listdir(directoryPath)
xTrain,yTrain,xTest,yTest = [],[],[]
try:
if len(dirList) < 1:
return None
imageDirPath = None
counter = 1
for directory in dirList:
imageDir = listdir('{}/{}'.format(directoryPath,directory))
splitPoint = int(SPLIT_POINT_COEFF * len(imageDir))
print('[{}] Loading dataset - {} images'.format(counter,directory))
counter += 1
trainImages,testimages = imageDir[:splitPoint],imageDir[splitPoint:]
imageDirPath = directoryPath + sep + directory + sep
self.addImagesToSet(imageDirPath,trainImages,directory,xTrain,yTrain)
self.addImagesToSet(imageDirPath,testimages,yTest)
except Exception as e:
print('Error: {}'.format(e))
return [],[]
return xTrain,yTest
def trainModel(self,trainDatasetDir):
train_dataset_dir = ('C:/Users/MUTHU/Documents/GitHub/Handwritten-math-symbols-recognition/Dataset')
print('Training.....')
xTrain,yTest = self.getTrainingAndTestData(trainDatasetDir)
if [] not in (xTrain,yTest):
randomForestClassifier = RandomForestClassifier()
randomForestClassifier.fit(xTrain,yTrain)
accuracyscore = randomForestClassifier.score(xTrain,yTrain)
# save classifier
pickle.dump(randomForestClassifier,open("Model/math_recognition_model.pkl",'wb'))
print("Model Accuracy score : {}".format(accuracyscore))
testAccuracyscore = randomForestClassifier.score(xTest,yTest)
print("Model Accuracy score (Test) : {}".format(testAccuracyscore))
else :
print("An error occurred.")
def predict(self,imagePath):
try:
image = [self.getimageVector(imagePath)]
# load saved model
try:
decisionTreeClassifierModel= pickle.load(open("Model/random_forest_classifier.pkl",'rb'))
modelPrediction = decisionTreeClassifierModel.predict(image)
print(modelPrediction)
print("Recognized expression:" + str(modelPrediction[0]))
except FileNotFoundError as modelFileError:
print("Error : {}".format(modelFileError))
self.trainModel(datasetDir)
self.predict(imagePath)
except FileNotFoundError as fileError:
print("Error : {}".format(fileError))
except Exception as e:
print("Error : {}".format(e))
解决方法
您需要在您的项目中单独导入它。为此，您可以参考以下链接： https://www.jetbrains.com/pycharm/guide/tips/install-and-import/
同时将您的pycharm更新到最新版本的2020.1.5，从菜单选项转到帮助->检查更新->选项将在下方右侧弹出以进行更新