신경망 활용 홈페이지 구축_홍승희_총코드_최종

[11기_홍승희]

■ 홈페이지 구축_홍승희_총코드

홈페이지 주소

https://hongseunghee.shinyapps.io/myapp27/

환경 구성 keras_study3에서 수행하였습니다

▩ [쉬움주의] 텐써 플로오 1.x 버전에서 keras 환경구성 

1. 환경구성

conda create -n keras_study3 

conda env  list 

activate  keras_study3 

conda install keras-base==2.3.1 

conda install tensorflow==1.14.0    

conda install spyder 

pip install matplotlib==3.0.3

1. 학습시키는 코드

"""지금까지 제일 잘 분류하는 신경망 코드 """

from keras.models import *

from keras.layers import  Dense, Flatten, Dropout, BatchNormalization, Conv2D, MaxPooling2D, Activation

from sklearn.model_selection import train_test_split

from keras.datasets.fashion_mnist import load_data

import matplotlib.pyplot as plt

import numpy as np

from keras.utils import to_categorical

from keras.layers.normalization import BatchNormalization

import loader3

from PIL import Image

"""이미지 경로 설정 """

train_image = 'c:\\animal2\\train_resize2\\'

test_image = 'c:\\animal2\\test_resize2\\'

train_label = 'c:\\animal2\\train_label.csv'

test_label = 'c:\\animal2\\test_label.csv'

""" loader3 함수를 가상환경 밑에 넣고 함수 구현 """

x_train = loader3.image_load(train_image)

y_train = loader3.label_load(train_label)

x_test = loader3.image_load(test_image)

y_test = loader3.label_load(test_label)      

print ( loader3.image_load(train_image).shape )

print ( loader3.image_load(test_image).shape )   # (400,)

print ( loader3.label_load(train_label).shape)   # (1600, 2)

print ( loader3.label_load(test_label).shape )

"""4차원 데이터이므로 reshape 필요없음"""

## x_train, x_test 수치 정규화를 위한 작업

x_mean = np.mean(x_train, axis = (0, 1, 2))

x_std = np.std(x_train, axis = (0, 1, 2))

x_train = (x_train - x_mean) / x_std

x_test = (x_test - x_mean) / x_std

# 검증 데이터셋을 만듭니다.

from sklearn.model_selection import train_test_split

np.random.seed(111)

# 훈련/테스트 데이터를 0.7/0.3의 비율로 분리합니다.

x_train, x_val, y_train, y_val = train_test_split(x_train, y_train,

                                                  test_size = 0.2, random_state = 777)

## 제너레이터 사용

from keras.preprocessing.image import ImageDataGenerator

train_datagen = ImageDataGenerator(horizontal_flip = True,

                                vertical_flip=True,

                                shear_range=0.5,

                                   width_shift_range = 0.1,

                                   height_shift_range = 0.1,

                                   rotation_range = 30,

                                   fill_mode = 'nearest')

val_datagen = ImageDataGenerator()

batch_size = 32

train_generator = train_datagen.flow(x_train, y_train,

                                    batch_size =32)

val_generator = val_datagen.flow(x_val, y_val,

                                batch_size =32)

from keras.models import *

from keras.layers import Conv2D, MaxPool2D, Dense, Flatten, BatchNormalization, Activation

from keras.optimizers import Adam

from keras.applications import VGG16

# imagenet을 학습한 모델을 불러옵니다.

vgg16 = VGG16(weights = 'imagenet', input_shape = (32, 32, 3), include_top = False)   

for layer in vgg16.layers[:-4]:

    layer.trainable = False

model = Sequential()    

model.add(vgg16)

model.add(Flatten())

model.add(Dense(256))

model.add(Activation('relu'))

model.add(Dense(128))

model.add(Activation('relu'))

model.add(Dense(64))

model.add(Activation('relu'))

model.add(Dense(2, activation = 'softmax'))

from keras.callbacks import ModelCheckpoint, EarlyStopping

filepath = 'c:\\animal2\\re_w2.h5'

callbacks = [ ModelCheckpoint( filepath = filepath , monitor = 'val_loss' , verbose=1, save_best_only=True ) ]

model.compile(optimizer = Adam(1e-4),

             loss = 'binary_crossentropy',

             metrics = ['acc'])

batch_size=32

def get_step(train_len, batch_size):      

    if(train_len % batch_size > 0):

        return train_len // batch_size + 1

    else:

        return train_len // batch_size

history = model.fit(train_generator,

                    epochs = 30,

                    steps_per_epoch = get_step(len(x_train), batch_size),

                    validation_data = val_generator,

                    validation_steps = get_step(len(x_val), batch_size), callbacks=callbacks )

print(model.eval‎uate(x_test,y_test))

save_model(model, 'c:\\animal2\\re_w3.h5')

2.R 스튜디오 코드 ( 홈페이지 구축용 )

server.R

ui.R

3. R 스튜디오 로컬용

setwd("D:\\yys267")

packages <- c('imager', "shiny", "jpeg", "png", "reticulate", "devtools")

if (length(setdiff(packages, rownames(installed.packages()))) > 0) {

  install.packages(setdiff(packages, rownames(installed.packages())))  

}

if (length(setdiff("keras", rownames(installed.packages()))) > 0) {

  devtools::install_github("rstudio/keras") 

}

require(imager)

require(shiny)

require(jpeg)

require(png)

library(reticulate)

library(keras)

#setwd(tempfile())      

#setwd("/Users/aiden/Desktop/data/cifar10_densenet")

load("envir.RData")

model<<-load_model_hdf5("final.h5")

synsets <<- readLines("synset.txt")

server <- shinyServer(function(input, output) {

  ntext <- eventReactive(input$goButton, {

    print(input$url)

    if (input$url == "http://") {

      NULL

    } else {

      tmp_file <- tempfile()

      download.file(input$url, destfile = tmp_file, mode = 'wb')

      tmp_file

    }

  })

  output$originImage = renderImage({

    list(src = if (input$tabs == "Upload Image") {

      if (is.null(input$file1)) {

        if (input$goButton == 0 || is.null(ntext())) {

          'jaguar.jpg'

        } else {

          ntext()

        }

      } else {

        input$file1$datapath

      }

    } else {

      if (input$goButton == 0 || is.null(ntext())) {

        if (is.null(input$file1)) {

          'jaguar.jpg'

        } else {

          input$file1$datapath

        }

      } else {

        ntext()

      }

    },

    title = "Original Image")

  }, deleteFile = FALSE)

  output$res <- renderText({

    src = if (input$tabs == "Upload Image") {

      if (is.null(input$file1)) {

        if (input$goButton == 0 || is.null(ntext())) {

          'jaguar.jpg'

        } else {

          ntext()

        }

      } else {

        input$file1$datapath

      }

    } else {

      if (input$goButton == 0 || is.null(ntext())) {

        if (is.null(input$file1)) {

          'jaguar.jpg'

        } else {

          input$file1$datapath

        }

      } else {

        ntext()

      }

    }

    img <- load.image(src)

    plot(img)

    img <- image_load(src, target_size = c(32,32))

    img

    x <- image_to_array(img)

    # ensure we have a 4d tensor with single element in the batch dimension,

    x <- array_reshape(x, c(1, dim(x)))

    # normalize

    x[,,,1] <- x[,,,1] /255.0

    x[,,,2] <- x[,,,2] /255.0

    x[,,,3] <- x[,,,3] /255.0

    # predcit

    preds <- model %>% predict(x)

    # output result as string

    order(preds[1,], decreasing = TRUE)

    max.idx <- order(preds[1,], decreasing = TRUE)[1]

    max.idx

    result <- synsets[max.idx]

    res_str <- ""

    tmp <- strsplit(result[1], " ")[[1]]

    res_str <- paste0(res_str, tmp[2])

    res_str

  })

})

require(imager)

require(shiny)

require(jpeg)

require(png)

ui <- shinyUI(

  fluidPage(

    includeCSS("bootstrap.css"),

    pageWithSidebar(

      headerPanel(title = '재규어 얼룩말 사진 판별 신경망',

                  windowTitle = 'Image Classification using DenseNet'),

      fluidRow(

        column(1),

        column(9,

               tabsetPanel(

                 id = "tabs",

                 tabPanel("Upload Image",

                          fileInput('file1', 'Upload a PNG / JPEG File:')),

                 tabPanel(

                   "Use the URL",

                   textInput("url", "Image URL:", "http://"),

                   actionButton("goButton", "Go!")

                 )

               ),

               h3(titlePanel("DESCRIPTION - Image Classification")),

               h3(titlePanel("Image Classification"))

        ),

        column(2)

      ),

      mainPanel(

        h3("Image"),

        tags$hr(),

        imageOutput("originImage", height = "auto"),

        tags$hr(),

        h3("What is this?"),

        tags$hr(),

        verbatimTextOutput("res")

      )

    )))

shinyApp(ui = ui, server = server)

[유연수]

고생했다. 승희야 ~