mxnet example 3 : PapSmear 2005 (Convoluted Neural Networks

The dataset is loaded from this work.

The images in the dataset illustrate different status of cervix cells. The images are classed by folder for each status. Each image focus on only one cell. There is two version of each image: original image with .BMP extension and filtered image with _d.bmp extension.

first of all, I separate these two kinds of images. The first step is to apply mxnet only on original images. for this I moved original image from carcinoma_in_situ/ folder to carcinoma_in_situ_org/, like this

# cd carcinoma_in_situ/
# mv `ls | grep '\w*.BMP'` ../carcinoma_in_situ_org/

In the end we have img_data filder which contains 7 subfolders. Each subfolder has no equal number of images. The number of images is indicated in the subfolder name.

ls img_data
## count all images
ls  img_data/* | wc -l

## carcinoma_in_situ_org_150
## carcinoma_in_situ_org_150_jpeg
## light_dysplastic_org_182_jpeg
## moderate_dysplastic_org_146_jpeg
## normal_columnar_org_98_jpeg
## normal_intermediate_org_70_jpeg
## normal_superficiel_org_74_jpeg
## severe_dysplastic_org_197_jpeg
##      934

Which format we have and is there the need to change it?

We will fellow this tutorial. The tutorial recommends RecordIOformat, which concatenates multiple examples into seekable binary files for better read efficiency. The image have .bmp format file whish is not appropriate for this tutorial. We need to convert the to pngor jpeg formats.

require(EBImage)

## Loading required package: EBImage

#dim(EBImage::readImage("img_data/carcinoma_in_situ_org_150/149143370-149143378-001.BMP"))

It seems that bmp format is not supported by EBImage. we try with imager package.

require(imager)

## Loading required package: imager

## Loading required package: magrittr

## 
## Attaching package: 'imager'

## The following object is masked from 'package:magrittr':
## 
##     add

## The following objects are masked from 'package:EBImage':
## 
##     channel, dilate, display, erode, resize, watershed

## The following objects are masked from 'package:stats':
## 
##     convolve, spectrum

## The following object is masked from 'package:graphics':
## 
##     frame

## The following object is masked from 'package:base':
## 
##     save.image

class(imager::load.image("img_data/carcinoma_in_situ_org_150/149143370-149143378-001.BMP"))

## [1] "cimg"         "imager_array" "numeric"

dim(imager::load.image("img_data/carcinoma_in_situ_org_150/149143370-149143378-001.BMP"))

## [1] 69 70  1  3

ok, We will use a loop to convert all images using imager load and save functions.

require(imager)
convert_image <- function(inputFolder, outputFolder, type = '.jpeg' ){
  
  names_files <- list.files(inputFolder)
  nfiles <- length(list.files(inputFolder))
  
  for(k in 1:nfiles){
    ## get image name
    nameFile <- names_files[k]
    # load image 
    tmp_img <- imager::load.image(paste0(inputFolder,nameFile, sep=""))
    # save image
    imager::save.image(im = tmp_img, file = paste0(outputFolder, tools::file_path_sans_ext(nameFile), type , sep =""),quality = 1)
  }
}
# convert_image(inputFolder = "img_data/carcinoma_in_situ_org_150/", outputFolder = "img_data/carcinoma_in_situ_org_150_jpeg/")
# convert_image("img_data/light_dysplastic_org_182/", "img_data/light_dysplastic_org_182_jpeg/")
# convert_image("img_data/moderate_dysplastic_org_146/", "img_data/moderate_dysplastic_org_146_jpeg/")
# convert_image("img_data/normal_columnar_org_98/", "img_data/normal_columnar_org_98_jpeg/")
# convert_image("img_data/normal_intermediate_org_70/", "img_data/normal_intermediate_org_70_jpeg/")
# convert_image("img_data/normal_superficiel_org_74/", "img_data/normal_superficiel_org_74_jpeg/")
# convert_image("img_data/severe_dysplastic_org_197/", "img_data/severe_dysplastic_org_197_jpeg/")

Now, We will use im2rec.py script to convert our images into RecordIO format.

convert image to RecorIO format (we will not use it)

We first prepare mydata.lst_train.lst and mydata.lst_val.lst files, which consist of the labels and image paths can be used for generating rec files. We need to create empty mydata.lst files before to run python script.

system('python im2rec.py --list --recursive --train-ratio 0.95 mydata.lst img_data')

The im2rec.py script generates the two files and get labels for each class of cells. If we look the number of images for train and val sets

cat mydata.lst_train.lst | wc -l
cat mydata.lst_val.lst | wc -l

##      878
##       47

The sum of the two listes gives 922 images (there are 8 images = 930 - 922 not found!). 47 corresponds to 5% of all images.

system('python im2rec.py --resize 480 --quality 95 --num-thread 16 mydata.lst img_data')

This command generates the mydata.lst_val.rec and mydata.lst_train.rec files. There are pre-setted models for mnist, cifar, and resnet datasets. But actually I don’t say how to use .rec dataset with theses models.

Come back to our jpeg image

## which image dimension will use. The images do not have the same dimension. We should to unify image dimension to 28x28 pixels.
## The use of bigger dimension, increase the time of the computing but not sure to have more accuracy. 

get_dim_im <- function(inputFolder ){
  #inputFolder <- 'img_data/carcinoma_in_situ_org_150_jpeg/'
  names_files <- list.files(inputFolder)
  nfiles <- length(list.files(inputFolder))
  ls_dim <- list()
  
  for(k in 1:nfiles){
    ## get image name
    nameFile <- names_files[k]
    # load image 
    tmp_img <- EBImage::readImage(paste0(inputFolder,nameFile, sep=""))
    
    #  Convert to grayscale
     tmp_img <- EBImage::channel(tmp_img, mode = 'gray')
     
    # Resize image to 28x28 pixels
    img_resized <- EBImage::resize(tmp_img, w = 28, h = 28)
    
    ## reduce the size of image 1/2
    ims <- EBImage::resize(tmp_img, dim(tmp_img)[1]/2)
    
    ls_dim[[k]] <- dim(ims)
  }
  plot(ims)
  plot(tmp_img)
  plot(img_resized)
  return(ls_dim[140:150])
}

get_dim_im("img_data/carcinoma_in_situ_org_150_jpeg/")

## [[1]]
## [1] 22 38
## 
## [[2]]
## [1] 44 44
## 
## [[3]]
## [1] 39 31
## 
## [[4]]
## [1] 67 60
## 
## [[5]]
## [1] 52 79
## 
## [[6]]
## [1] 52 64
## 
## [[7]]
## [1] 36 51
## 
## [[8]]
## [1] 41 43
## 
## [[9]]
## [1] 33 48
## 
## [[10]]
## [1] 41 41
## 
## [[11]]
## [1] 37 40

Convert images from a folder to matrix: Each image in one row, set dimension 28x28x1.

#source("../R/rbindna.R")
require(EBImage)
images2matrix <- function(inputFolder, w= 28, h = 28, class){
  
  names_files <- list.files(inputFolder)
  nfiles <- length(list.files(inputFolder))
  
  df <- data.frame(matrix(ncol = (w*h)+1, nrow = 0))
  # Set names. The first column is the labels, the other columns are the pixels.
  colnames(df) <-  c("Labels", paste("pixel", c(1:(w*h) )))

  
  for(k in 1:nfiles){
    ## get image name
    nameFile <- names_files[k]
    # load image 
    tmp_img <- EBImage::readImage(paste0(inputFolder,nameFile, sep=""))
    
    ## convet to grayscale
    tmp_img <- EBImage::channel(tmp_img, mode = 'gray')
    
     # Resize image to 28x28 pixels
    ims <- EBImage::resize(tmp_img, w = w, h = h)
    ## reduce the size of image 1/2
    #ims <- EBImage::resize(tmp_img, dim(tmp_img)[1] * scale)
    
    # Get image matrix (there should be another function to do this faster and more neatly!)
    img_matrix <- ims@.Data
    #print(dim(img_matrix))
    
    # Coerce to a vector
    img_vector <- as.vector(img_matrix)
  
    
    # Add label
    label <- class
    vec <- c(label, img_vector)
    # Stack in rs_df using rbind
    ## AVoid rbind, it ignore empty dataframe and colnames
    #df <- rbind(df_bkp, vec)
     
     df[nrow(df)+1,] <- vec
    
     # Print status
     print(paste(k,names_files[k],sep = " "))
    
  }
  
  rownames(df) <- NULL
  return(df)
  
}

class1_mat <- images2matrix("img_data/normal_intermediate_org_70_jpeg/", w=28, h=28 ,class= "1")

## [1] "1 153955676-153955721-001.jpeg"
## [1] "2 153955676-153955746-001.jpeg"
## [1] "3 153956279-153956296-001.jpeg"
## [1] "4 153958547-153958621-001.jpeg"
## [1] "5 157183332-157183388-001.jpeg"
## [1] "6 157183332-157183388-002.jpeg"
## [1] "7 157183412-157183656-001.jpeg"
## [1] "8 157183828-157183840-001.jpeg"
## [1] "9 157183828-157183877-001.jpeg"
## [1] "10 157183828-157183887-001.jpeg"
## [1] "11 157184013-157184023-001.jpeg"
## [1] "12 157184013-157184039-001.jpeg"
## [1] "13 157184067-157184080-001.jpeg"
## [1] "14 157184112-157184120-001.jpeg"
## [1] "15 157185575-157185586-001.jpeg"
## [1] "16 157185677-157185690-001.jpeg"
## [1] "17 157222534-157222561-001.jpeg"
## [1] "18 157222534-157222579-001.jpeg"
## [1] "19 157222534-157223349-001.jpeg"
## [1] "20 157222534-157223355-001.jpeg"
## [1] "21 157222534-157223364-001.jpeg"
## [1] "22 157222647-157222660-001.jpeg"
## [1] "23 157222647-157222687-001.jpeg"
## [1] "24 157222801-157222811-001.jpeg"
## [1] "25 157222801-157222811-002.jpeg"
## [1] "26 157222801-157222844-001.jpeg"
## [1] "27 157223246-157223258-001.jpeg"
## [1] "28 157223321-157223328-001.jpeg"
## [1] "29 157223321-157223341-001.jpeg"
## [1] "30 157223394-157223406-001.jpeg"
## [1] "31 157223394-157223406-002.jpeg"
## [1] "32 157223394-157223428-001.jpeg"
## [1] "33 157223394-157223428-002.jpeg"
## [1] "34 157223549-157223570-001.jpeg"
## [1] "35 157223779-157224140-001.jpeg"
## [1] "36 157224458-157224464-001.jpeg"
## [1] "37 157268504-157268534-001.jpeg"
## [1] "38 158989297-158989324-001.jpeg"
## [1] "39 209047526-209047626-001.jpeg"
## [1] "40 209047526-209047762-001.jpeg"
## [1] "41 209047881-209047928-001.jpeg"
## [1] "42 209048086-209048239-001.jpeg"
## [1] "43 209307421-209307504-001.jpeg"
## [1] "44 209307421-209307561-001.jpeg"
## [1] "45 209522316-209522391-001.jpeg"
## [1] "46 209522474-209522522-001.jpeg"
## [1] "47 209522474-209522554-001.jpeg"
## [1] "48 209522641-209522674-001.jpeg"
## [1] "49 209522800-209522835-001.jpeg"
## [1] "50 209522940-209522970-001.jpeg"
## [1] "51 209522940-209522991-001.jpeg"
## [1] "52 209522940-209523052-001.jpeg"
## [1] "53 209565409-209565466-001.jpeg"
## [1] "54 209565409-209565503-001.jpeg"
## [1] "55 209565409-209565600-001.jpeg"
## [1] "56 209565698-209565729-001.jpeg"
## [1] "57 209565698-209565772-001.jpeg"
## [1] "58 209565864-209565890-001.jpeg"
## [1] "59 209565864-209565911-001.jpeg"
## [1] "60 209565864-209565950-001.jpeg"
## [1] "61 209566047-209566095-001.jpeg"
## [1] "62 209566047-209566125-001.jpeg"
## [1] "63 209566205-209566247-001.jpeg"
## [1] "64 209566205-209566266-001.jpeg"
## [1] "65 209566205-209566289-001.jpeg"
## [1] "66 209566205-209566321-001.jpeg"
## [1] "67 209566205-209566333-001.jpeg"
## [1] "68 209566399-209566464-001.jpeg"
## [1] "69 209566399-209566485-001.jpeg"
## [1] "70 209566399-209566517-001.jpeg"

 class2_mat <- images2matrix("img_data/normal_columnar_org_98_jpeg/", w=28, h=28 , class= '2')

## [1] "1 153956040-153956058-001.jpeg"
## [1] "2 153956040-153956058-002.jpeg"
## [1] "3 153956040-153956058-003.jpeg"
## [1] "4 153956040-153956058-004.jpeg"
## [1] "5 153956040-153956058-005.jpeg"
## [1] "6 153956040-153956058-006.jpeg"
## [1] "7 153956040-153956072-001.jpeg"
## [1] "8 153956040-153956072-002.jpeg"
## [1] "9 153956040-153956072-003.jpeg"
## [1] "10 153956040-153956072-004.jpeg"
## [1] "11 153956444-153956458-001.jpeg"
## [1] "12 153956444-153956458-002.jpeg"
## [1] "13 153956444-153956458-003.jpeg"
## [1] "14 153956444-153956458-004.jpeg"
## [1] "15 153958154-153958168-001.jpeg"
## [1] "16 153958154-153958168-003.jpeg"
## [1] "17 153958154-153958179-001.jpeg"
## [1] "18 153958154-153958179-002.jpeg"
## [1] "19 153958154-153958194-001.jpeg"
## [1] "20 153958154-153958194-002.jpeg"
## [1] "21 153958154-153958194-003.jpeg"
## [1] "22 153958154-153958194-004.jpeg"
## [1] "23 153958154-153958211-001.jpeg"
## [1] "24 153958154-153958248-002.jpeg"
## [1] "25 153958154-153958248-003.jpeg"
## [1] "26 153958345-153958364-001.jpeg"
## [1] "27 153958547-153958572-001.jpeg"
## [1] "28 153958547-153958572-002.jpeg"
## [1] "29 153958547-153958572-003.jpeg"
## [1] "30 153958547-153958572-004.jpeg"
## [1] "31 153958547-153958572-005.jpeg"
## [1] "32 153958547-153958572-006.jpeg"
## [1] "33 153958547-153958572-007.jpeg"
## [1] "34 153958547-153958572-008.jpeg"
## [1] "35 153958547-153958572-009.jpeg"
## [1] "36 153958547-153958590-001.jpeg"
## [1] "37 153958547-153958590-002.jpeg"
## [1] "38 157181146-157181197-001.jpeg"
## [1] "39 157181146-157181197-002.jpeg"
## [1] "40 157181146-157181197-003.jpeg"
## [1] "41 157181481-157181497-001.jpeg"
## [1] "42 157181481-157181497-002.jpeg"
## [1] "43 157181481-157181497-003.jpeg"
## [1] "44 157181481-157181497-004.jpeg"
## [1] "45 157184850-157184920-001.jpeg"
## [1] "46 157184850-157184920-002.jpeg"
## [1] "47 157185433-157185454-001.jpeg"
## [1] "48 157185433-157185479-001.jpeg"
## [1] "49 157185433-157185479-002.jpeg"
## [1] "50 157185433-157185479-003.jpeg"
## [1] "51 157185433-157185479-004.jpeg"
## [1] "52 157185433-157185479-005.jpeg"
## [1] "53 157185433-157185479-006.jpeg"
## [1] "54 157185433-157185479-007.jpeg"
## [1] "55 157185433-157185508-001.jpeg"
## [1] "56 157185433-157185508-002.jpeg"
## [1] "57 157185433-157185508-003.jpeg"
## [1] "58 157185433-157185508-004.jpeg"
## [1] "59 157185433-157185508-005.jpeg"
## [1] "60 157185433-157185508-006.jpeg"
## [1] "61 157185527-157185540-001.jpeg"
## [1] "62 157185527-157185540-002.jpeg"
## [1] "63 157185527-157185540-003.jpeg"
## [1] "64 157185527-157185540-004.jpeg"
## [1] "65 157185730-157185758-001.jpeg"
## [1] "66 157185730-157185758-002.jpeg"
## [1] "67 157185781-157185793-001.jpeg"
## [1] "68 157185781-157185793-002.jpeg"
## [1] "69 157185781-157185793-003.jpeg"
## [1] "70 157185781-157185814-001.jpeg"
## [1] "71 157185781-157185814-002.jpeg"
## [1] "72 157224172-157224207-001.jpeg"
## [1] "73 157224172-157224207-002.jpeg"
## [1] "74 157224172-157224207-003.jpeg"
## [1] "75 157227058-157227087-001.jpeg"
## [1] "76 157227461-157227503-001.jpeg"
## [1] "77 157227461-157227503-002.jpeg"
## [1] "78 157227461-157227503-003.jpeg"
## [1] "79 157266930-157266947-001.jpeg"
## [1] "80 157266930-157266947-002.jpeg"
## [1] "81 157266930-157266947-003.jpeg"
## [1] "82 157267001-157267013-001.jpeg"
## [1] "83 157267059-157267072-001.jpeg"
## [1] "84 157267059-157267072-002.jpeg"
## [1] "85 157267059-157267072-003.jpeg"
## [1] "86 157267059-157267072-004.jpeg"
## [1] "87 157267263-157267286-001.jpeg"
## [1] "88 157267263-157267286-002.jpeg"
## [1] "89 158986766-158986776-001.jpeg"
## [1] "90 158986766-158986776-002.jpeg"
## [1] "91 158986813-158986820-001.jpeg"
## [1] "92 158986813-158986820-002.jpeg"
## [1] "93 158986920-158986928-001.jpeg"
## [1] "94 158986920-158986928-002.jpeg"
## [1] "95 158986920-158986928-003.jpeg"
## [1] "96 158986920-158986928-004.jpeg"
## [1] "97 158986920-158986928-005.jpeg"
## [1] "98 158986920-158986928-006.jpeg"

 class3_mat <- images2matrix("img_data/normal_superficiel_org_74_jpeg/", w=28, h=28 , class= '3')

## [1] "1 153958345-153958392-001.jpeg"
## [1] "2 153960256-153960295-002.jpeg"
## [1] "3 157181281-157181308-001.jpeg"
## [1] "4 157181387-157181418-001.jpeg"
## [1] "5 157181525-157181540-001.jpeg"
## [1] "6 157181569-157181599-001.jpeg"
## [1] "7 157181671-157181686-001.jpeg"
## [1] "8 157181671-157181697-001.jpeg"
## [1] "9 157181671-157181708-001.jpeg"
## [1] "10 157183193-157183241-001.jpeg"
## [1] "11 157183193-157183279-001.jpeg"
## [1] "12 157183332-157183346-001.jpeg"
## [1] "13 157183332-157183361-001.jpeg"
## [1] "14 157183332-157183376-001.jpeg"
## [1] "15 157183412-157183614-001.jpeg"
## [1] "16 157183412-157183638-001.jpeg"
## [1] "17 157183722-157183783-001.jpeg"
## [1] "18 157183722-157183783-002.jpeg"
## [1] "19 157183964-157183980-001.jpeg"
## [1] "20 157183964-157183991-001.jpeg"
## [1] "21 157184013-157184031-001.jpeg"
## [1] "22 157184112-157184128-001.jpeg"
## [1] "23 157184850-157184870-001.jpeg"
## [1] "24 157184850-157184891-001.jpeg"
## [1] "25 157184850-157184942-001.jpeg"
## [1] "26 157184850-157184976-001.jpeg"
## [1] "27 157184999-157185032-001.jpeg"
## [1] "28 157185054-157185079-001.jpeg"
## [1] "29 157185133-157185143-001.jpeg"
## [1] "30 157185164-157185177-001.jpeg"
## [1] "31 157185164-157185196-001.jpeg"
## [1] "32 157185208-157185238-001.jpeg"
## [1] "33 157185309-157185360-001.jpeg"
## [1] "34 157185613-157185627-001.jpeg"
## [1] "35 157222534-157222607-001.jpeg"
## [1] "36 157222737-157222750-001.jpeg"
## [1] "37 157222737-157222750-002.jpeg"
## [1] "38 157222904-157223196-001.jpeg"
## [1] "39 157222904-157223217-001.jpeg"
## [1] "40 157223659-157223667-001.jpeg"
## [1] "41 157223659-157223667-002.jpeg"
## [1] "42 157223735-157223751-001.jpeg"
## [1] "43 157223735-157223766-001.jpeg"
## [1] "44 157223779-157224126-001.jpeg"
## [1] "45 157224172-157224179-001.jpeg"
## [1] "46 157224244-157224258-001.jpeg"
## [1] "47 157224297-157224304-001.jpeg"
## [1] "48 157224297-157224304-002.jpeg"
## [1] "49 157224297-157224320-001.jpeg"
## [1] "50 157224297-157224320-002.jpeg"
## [1] "51 157224297-157224336-001.jpeg"
## [1] "52 157224352-157224395-001.jpeg"
## [1] "53 157224412-157224429-001.jpeg"
## [1] "54 157224458-157224483-001.jpeg"
## [1] "55 157224504-157224520-001.jpeg"
## [1] "56 157267647-157267732-001.jpeg"
## [1] "57 157268242-157268296-001.jpeg"
## [1] "58 157268342-157268376-001.jpeg"
## [1] "59 157268342-157268401-001.jpeg"
## [1] "60 157268504-157268544-001.jpeg"
## [1] "61 157268587-157268617-001.jpeg"
## [1] "62 158987033-158987057-001.jpeg"
## [1] "63 158987453-158987462-001.jpeg"
## [1] "64 158987493-158987499-001.jpeg"
## [1] "65 158987493-158987505-001.jpeg"
## [1] "66 209047342-209047400-001.jpeg"
## [1] "67 209047342-209047443-001.jpeg"
## [1] "68 209047342-209047478-001.jpeg"
## [1] "69 209047526-209047717-001.jpeg"
## [1] "70 209047526-209047798-001.jpeg"
## [1] "71 209047881-209048017-001.jpeg"
## [1] "72 209048086-209048137-001.jpeg"
## [1] "73 209048086-209048278-001.jpeg"
## [1] "74 209307421-209307597-001.jpeg"

# class4_mat <- images2matrix("img_data/light_dysplastic_org_182_jpeg/", w=28, h=28 , class= '4')
# class5_mat <- images2matrix("img_data/moderate_dysplastic_org_146_jpeg/", w=28, h=28 , class= '5')
# class6_mat <- images2matrix("img_data/carcinoma_in_situ_org_150_jpeg/", w=28, h=28 , class= '6')
# class7_mat <- images2matrix("img_data/severe_dysplastic_org_197_jpeg/", w=28, h=28 , class= '7')
listFiles <- list.files("img_data",full.names = TRUE)[-1]
classes <- list()
for(i in seq_len(length(listFiles))){
  
  classes[[paste0("c",i)]] <- images2matrix(paste0(listFiles[i], "/"), w= 28, h= 28, class= i)
}

## [1] "1 149143370-149143378-001.jpeg"
## [1] "2 149143370-149143378-002.jpeg"
## [1] "3 149143370-149143378-003.jpeg"
## [1] "4 149143370-149143388-001.jpeg"
## [1] "5 149143370-149143388-002.jpeg"
## [1] "6 149143370-149143388-003.jpeg"
## [1] "7 149146635-149146651-001.jpeg"
## [1] "8 149146635-149146651-002.jpeg"
## [1] "9 149146635-149146651-003.jpeg"
## [1] "10 149146837-149146854-001.jpeg"
## [1] "11 149146886-149146936-001.jpeg"
## [1] "12 149146886-149146936-002.jpeg"
## [1] "13 149147060-149147075-001.jpeg"
## [1] "14 149147060-149147075-002.jpeg"
## [1] "15 149147060-149147075-003.jpeg"
## [1] "16 149147848-149147864-001.jpeg"
## [1] "17 149147848-149147864-002.jpeg"
## [1] "18 149147848-149147864-003.jpeg"
## [1] "19 149147949-149147960-001.jpeg"
## [1] "20 149147949-149147960-002.jpeg"
## [1] "21 149148124-149148168-001.jpeg"
## [1] "22 149148124-149148181-001.jpeg"
## [1] "23 149148124-149148181-002.jpeg"
## [1] "24 149148124-149148181-003.jpeg"
## [1] "25 149148124-149148181-004.jpeg"
## [1] "26 149148669-149148707-001.jpeg"
## [1] "27 149148669-149148707-002.jpeg"
## [1] "28 149148669-149148707-003.jpeg"
## [1] "29 149181904-149181927-001.jpeg"
## [1] "30 149181904-149181951-001.jpeg"
## [1] "31 149181904-149181951-002.jpeg"
## [1] "32 149181904-149182017-001.jpeg"
## [1] "33 149181904-149182017-002.jpeg"
## [1] "34 149181904-149182017-003.jpeg"
## [1] "35 149181904-149182044-001.jpeg"
## [1] "36 149181904-149182044-002.jpeg"
## [1] "37 149181904-149182044-003.jpeg"
## [1] "38 149182125-149182135-001.jpeg"
## [1] "39 149182125-149182135-002.jpeg"
## [1] "40 149182125-149182135-003.jpeg"
## [1] "41 149182355-149182396-001.jpeg"
## [1] "42 149182550-149182571-001.jpeg"
## [1] "43 149182550-149182571-002.jpeg"
## [1] "44 149182550-149182571-003.jpeg"
## [1] "45 149182550-149182571-004.jpeg"
## [1] "46 149182550-149182608-001.jpeg"
## [1] "47 149182550-149182608-002.jpeg"
## [1] "48 149182550-149182608-003.jpeg"
## [1] "49 149182550-149182608-004.jpeg"
## [1] "50 149182550-149182608-005.jpeg"
## [1] "51 149182657-149182667-001.jpeg"
## [1] "52 149182657-149182667-002.jpeg"
## [1] "53 149182657-149182667-003.jpeg"
## [1] "54 149182657-149182681-001.jpeg"
## [1] "55 149182657-149182681-002.jpeg"
## [1] "56 149182657-149182681-003.jpeg"
## [1] "57 149182657-149182698-001.jpeg"
## [1] "58 149182657-149182698-002.jpeg"
## [1] "59 149182657-149182698-003.jpeg"
## [1] "60 149182657-149182713-001.jpeg"
## [1] "61 149182657-149182713-002.jpeg"
## [1] "62 149182657-149182713-003.jpeg"
## [1] "63 149185405-149185423-001.jpeg"
## [1] "64 149185405-149185423-002.jpeg"
## [1] "65 149185405-149185457-001.jpeg"
## [1] "66 149185497-149185527-001.jpeg"
## [1] "67 149185497-149185527-002.jpeg"
## [1] "68 149185497-149185527-003.jpeg"
## [1] "69 149185604-149185616-001.jpeg"
## [1] "70 149185604-149185629-001.jpeg"
## [1] "71 149185604-149185645-001.jpeg"
## [1] "72 149185604-149185645-002.jpeg"
## [1] "73 149185604-149185645-003.jpeg"
## [1] "74 149185604-149185645-004.jpeg"
## [1] "75 149185604-149185645-005.jpeg"
## [1] "76 149185604-149185657-001.jpeg"
## [1] "77 149185604-149185657-002.jpeg"
## [1] "78 149185698-149185709-001.jpeg"
## [1] "79 149185698-149185709-002.jpeg"
## [1] "80 149185698-149185721-001.jpeg"
## [1] "81 149185698-149185721-002.jpeg"
## [1] "82 149185698-149185721-003.jpeg"
## [1] "83 149185698-149185721-004.jpeg"
## [1] "84 149185698-149185744-001.jpeg"
## [1] "85 149185788-149185824-001.jpeg"
## [1] "86 149185850-149185864-001.jpeg"
## [1] "87 149185850-149185875-001.jpeg"
## [1] "88 149187049-149187073-001.jpeg"
## [1] "89 149187049-149187090-001.jpeg"
## [1] "90 149187049-149187100-002.jpeg"
## [1] "91 149314453-149314465-001.jpeg"
## [1] "92 149314453-149314507-001.jpeg"
## [1] "93 149314453-149314507-002.jpeg"
## [1] "94 149314453-149314507-003.jpeg"
## [1] "95 149314453-149314507-004.jpeg"
## [1] "96 149314679-149314698-001.jpeg"
## [1] "97 149314679-149314698-002.jpeg"
## [1] "98 149314679-149314698-003.jpeg"
## [1] "99 149314753-149314760-001.jpeg"
## [1] "100 149314881-149314903-001.jpeg"
## [1] "101 149315229-149315249-001.jpeg"
## [1] "102 149315229-149315249-002.jpeg"
## [1] "103 149315372-149315390-001.jpeg"
## [1] "104 149315671-149315731-001.jpeg"
## [1] "105 149315671-149315731-002.jpeg"
## [1] "106 149315671-149315740-001.jpeg"
## [1] "107 149315671-149315740-002.jpeg"
## [1] "108 149315671-149315740-003.jpeg"
## [1] "109 149315671-149315740-004.jpeg"
## [1] "110 149315671-149315749-001.jpeg"
## [1] "111 149315671-149315749-002.jpeg"
## [1] "112 149315775-149315790-001.jpeg"
## [1] "113 149315775-149315790-002.jpeg"
## [1] "114 149315775-149315790-003.jpeg"
## [1] "115 149316025-149316032-001.jpeg"
## [1] "116 149316117-149316122-002.jpeg"
## [1] "117 149316117-149316122-003.jpeg"
## [1] "118 149316117-149316122-004.jpeg"
## [1] "119 149316117-149316131-001.jpeg"
## [1] "120 149316117-149316131-002.jpeg"
## [1] "121 149316426-149316462-001.jpeg"
## [1] "122 149316426-149316462-002.jpeg"
## [1] "123 149316426-149316462-003.jpeg"
## [1] "124 149316426-149316462-004.jpeg"
## [1] "125 149317114-149317152-001.jpeg"
## [1] "126 149317114-149317152-002.jpeg"
## [1] "127 149317114-149317152-004.jpeg"
## [1] "128 149357956-149358043-001.jpeg"
## [1] "129 153826597-153826619-001.jpeg"
## [1] "130 153826597-153826619-002.jpeg"
## [1] "131 153827595-153827657-001.jpeg"
## [1] "132 153827595-153827657-002.jpeg"
## [1] "133 153827595-153827664-001.jpeg"
## [1] "134 153827595-153827669-001.jpeg"
## [1] "135 153830680-153830693-001.jpeg"
## [1] "136 153831027-153831036-001.jpeg"
## [1] "137 153831027-153831036-002.jpeg"
## [1] "138 153831027-153831045-001.jpeg"
## [1] "139 153831027-153831045-002.jpeg"
## [1] "140 153831027-153831045-003.jpeg"
## [1] "141 153831426-153831438-001.jpeg"
## [1] "142 153831471-153831479-001.jpeg"
## [1] "143 153831471-153831486-001.jpeg"
## [1] "144 153831471-153831486-002.jpeg"
## [1] "145 153831471-153831486-003.jpeg"
## [1] "146 153831471-153831486-004.jpeg"
## [1] "147 153916114-153916150-001.jpeg"
## [1] "148 153916114-153916167-001.jpeg"
## [1] "149 153916114-153916167-002.jpeg"
## [1] "150 153916569-153916586-001.jpeg"
## [1] "1 148494967-148494986-001.jpeg"
## [1] "2 148495491-148495504-001.jpeg"
## [1] "3 148495553-148495585-001.jpeg"
## [1] "4 148495842-148495854-001.jpeg"
## [1] "5 148495842-148495863-001.jpeg"
## [1] "6 148495885-148495896-001.jpeg"
## [1] "7 148497551-148497588-001.jpeg"
## [1] "8 148497551-148497625-001.jpeg"
## [1] "9 148497676-148497704-001.jpeg"
## [1] "10 148497676-148497718-001.jpeg"
## [1] "11 148497788-148497816-001.jpeg"
## [1] "12 148497788-148497816-002.jpeg"
## [1] "13 148497788-148497836-001.jpeg"
## [1] "14 148497788-148497836-002.jpeg"
## [1] "15 148498683-148498703-001.jpeg"
## [1] "16 148498683-148498716-001.jpeg"
## [1] "17 148499274-148499302-001.jpeg"
## [1] "18 148499383-148499424-002.jpeg"
## [1] "19 148499383-148499452-001.jpeg"
## [1] "20 148499383-148499452-002.jpeg"
## [1] "21 148499383-148499474-001.jpeg"
## [1] "22 148499815-148500094-001.jpeg"
## [1] "23 148500154-148500165-001.jpeg"
## [1] "24 148503071-148503092-001.jpeg"
## [1] "25 148503071-148503092-002.jpeg"
## [1] "26 148503209-148503222-001.jpeg"
## [1] "27 148503209-148503222-003.jpeg"
## [1] "28 148503209-148503231-001.jpeg"
## [1] "29 148503400-148503415-003.jpeg"
## [1] "30 148503400-148503424-001.jpeg"
## [1] "31 148503400-148503432-001.jpeg"
## [1] "32 148503568-148503580-001.jpeg"
## [1] "33 148503568-148503580-002.jpeg"
## [1] "34 148503568-148503580-003.jpeg"
## [1] "35 148503568-148503598-001.jpeg"
## [1] "36 148503664-148503675-001.jpeg"
## [1] "37 148503664-148503675-002.jpeg"
## [1] "38 148503664-148503685-001.jpeg"
## [1] "39 148503664-148503685-002.jpeg"
## [1] "40 148503900-148503914-001.jpeg"
## [1] "41 148710162-148710195-001.jpeg"
## [1] "42 148710162-148710195-002.jpeg"
## [1] "43 148714237-148714247-001.jpeg"
## [1] "44 148714902-148714917-001.jpeg"
## [1] "45 148718455-148718461-001.jpeg"
## [1] "46 148719308-148719348-002.jpeg"
## [1] "47 148719308-148719348-004.jpeg"
## [1] "48 148719308-148719348-005.jpeg"
## [1] "49 148719308-148719357-001.jpeg"
## [1] "50 148719308-148719357-002.jpeg"
## [1] "51 148719372-148719378-001.jpeg"
## [1] "52 148839765-148839773-001.jpeg"
## [1] "53 148839825-148839835-001.jpeg"
## [1] "54 148841826-148841854-001.jpeg"
## [1] "55 148842055-148842064-001.jpeg"
## [1] "56 148881675-148881704-001.jpeg"
## [1] "57 148881729-148881744-001.jpeg"
## [1] "58 148881870-148881960-001.jpeg"
## [1] "59 148881870-148881960-002.jpeg"
## [1] "60 148881870-148882178-001.jpeg"
## [1] "61 148881870-148882178-002.jpeg"
## [1] "62 148881870-148882187-001.jpeg"
## [1] "63 148882307-148882323-001.jpeg"
## [1] "64 148882307-148882323-002.jpeg"
## [1] "65 148882307-148882410-001.jpeg"
## [1] "66 148882307-148882447-001.jpeg"
## [1] "67 148882535-148882616-001.jpeg"
## [1] "68 148882535-148882616-002.jpeg"
## [1] "69 148882671-148882687-001.jpeg"
## [1] "70 148882671-148882687-002.jpeg"
## [1] "71 148883900-148883907-001.jpeg"
## [1] "72 148883996-148884017-001.jpeg"
## [1] "73 148884702-148884710-001.jpeg"
## [1] "74 148884702-148884710-002.jpeg"
## [1] "75 148886869-148886897-001.jpeg"
## [1] "76 149011613-149011656-001.jpeg"
## [1] "77 149014929-149015022-001.jpeg"
## [1] "78 149014929-149015022-002.jpeg"
## [1] "79 149015418-149015435-001.jpeg"
## [1] "80 149015418-149015435-002.jpeg"
## [1] "81 149053423-149053447-001.jpeg"
## [1] "82 149053945-149053983-001.jpeg"
## [1] "83 149054277-149054304-001.jpeg"
## [1] "84 149054277-149054304-002.jpeg"
## [1] "85 149054526-149054540-001.jpeg"
## [1] "86 149054526-149054563-001.jpeg"
## [1] "87 149054642-149054677-002.jpeg"
## [1] "88 149054642-149054677-003.jpeg"
## [1] "89 149058881-149058910-001.jpeg"
## [1] "90 149059215-149059228-001.jpeg"
## [1] "91 149059215-149059228-002.jpeg"
## [1] "92 149061666-149061726-001.jpeg"
## [1] "93 149061666-149061726-002.jpeg"
## [1] "94 149061666-149061726-003.jpeg"
## [1] "95 149095913-149095926-001.jpeg"
## [1] "96 149095972-149095990-001.jpeg"
## [1] "97 149099276-149099290-001.jpeg"
## [1] "98 149099276-149099290-002.jpeg"
## [1] "99 149099398-149099421-001.jpeg"
## [1] "100 149099502-149099525-001.jpeg"
## [1] "101 149099502-149099525-002.jpeg"
## [1] "102 149100006-149100018-001.jpeg"
## [1] "103 149102442-149102451-001.jpeg"
## [1] "104 149102442-149102460-002.jpeg"
## [1] "105 149105003-149105050-001.jpeg"
## [1] "106 149105003-149105050-002.jpeg"
## [1] "107 149140169-149140178-001.jpeg"
## [1] "108 149140169-149140178-002.jpeg"
## [1] "109 149140169-149140188-001.jpeg"
## [1] "110 149316852-149316872-001.jpeg"
## [1] "111 149317002-149317018-001.jpeg"
## [1] "112 149357849-149357874-001.jpeg"
## [1] "113 149788716-149788727-001.jpeg"
## [1] "114 149788716-149788727-002.jpeg"
## [1] "115 149788716-149788727-003.jpeg"
## [1] "116 149788987-149788996-001.jpeg"
## [1] "117 149788987-149788996-003.jpeg"
## [1] "118 149788987-149789006-001.jpeg"
## [1] "119 153266866-153266917-001.jpeg"
## [1] "120 153266866-153266917-002.jpeg"
## [1] "121 153268036-153268083-001.jpeg"
## [1] "122 153268807-153268826-001.jpeg"
## [1] "123 153275719-153275737-001.jpeg"
## [1] "124 153275831-153275844-001.jpeg"
## [1] "125 153313074-153313079-001.jpeg"
## [1] "126 153314083-153314092-002.jpeg"
## [1] "127 153314466-153314473-001.jpeg"
## [1] "128 153314466-153314473-002.jpeg"
## [1] "129 153314466-153314473-003.jpeg"
## [1] "130 153314759-153314767-002.jpeg"
## [1] "131 153314759-153314767-003.jpeg"
## [1] "132 153314857-153314882-001.jpeg"
## [1] "133 153314857-153314882-002.jpeg"
## [1] "134 153354589-153354597-001.jpeg"
## [1] "135 153354589-153354597-002.jpeg"
## [1] "136 153354858-153354901-001.jpeg"
## [1] "137 153354858-153354901-002.jpeg"
## [1] "138 153655593-153655600-001.jpeg"
## [1] "139 153655593-153655600-002.jpeg"
## [1] "140 153655593-153655608-001.jpeg"
## [1] "141 153656862-153656873-001.jpeg"
## [1] "142 153656862-153656891-001.jpeg"
## [1] "143 153656957-153656964-001.jpeg"
## [1] "144 153657257-153657289-001.jpeg"
## [1] "145 153657257-153657289-002.jpeg"
## [1] "146 153657327-153657363-001.jpeg"
## [1] "147 153657327-153657363-002.jpeg"
## [1] "148 153657419-153657488-001.jpeg"
## [1] "149 153657419-153657488-002.jpeg"
## [1] "150 153657599-153657610-001.jpeg"
## [1] "151 153657599-153657610-002.jpeg"
## [1] "152 153657599-153657622-002.jpeg"
## [1] "153 153657599-153657633-001.jpeg"
## [1] "154 153657599-153657633-002.jpeg"
## [1] "155 153657698-153657708-001.jpeg"
## [1] "156 153697097-153697106-001.jpeg"
## [1] "157 153697726-153697736-001.jpeg"
## [1] "158 153700207-153700215-001.jpeg"
## [1] "159 153700207-153700224-001.jpeg"
## [1] "160 153701009-153701019-001.jpeg"
## [1] "161 153701009-153701019-002.jpeg"
## [1] "162 153701139-153701148-001.jpeg"
## [1] "163 153701557-153701566-001.jpeg"
## [1] "164 153701949-153701958-001.jpeg"
## [1] "165 153701949-153701964-001.jpeg"
## [1] "166 153702037-153702051-001.jpeg"
## [1] "167 153702037-153702051-002.jpeg"
## [1] "168 153702037-153702051-003.jpeg"
## [1] "169 153702037-153702060-001.jpeg"
## [1] "170 153829664-153829672-001.jpeg"
## [1] "171 153829700-153829705-001.jpeg"
## [1] "172 153829745-153829754-001.jpeg"
## [1] "173 154519964-154519981-001.jpeg"
## [1] "174 154520056-154520096-001.jpeg"
## [1] "175 204870858-204870872-001.jpeg"
## [1] "176 204870858-204870872-002.jpeg"
## [1] "177 204870926-204870933-001.jpeg"
## [1] "178 204870926-204870933-002.jpeg"
## [1] "179 204870951-204870961-001.jpeg"
## [1] "180 204870951-204870961-002.jpeg"
## [1] "181 204871030-204871038-001.jpeg"
## [1] "182 204871030-204871038-002.jpeg"
## [1] "1 148848523-148848538-001.jpeg"
## [1] "2 148848523-148848559-001.jpeg"
## [1] "3 148848523-148848559-002.jpeg"
## [1] "4 148848523-148848559-003.jpeg"
## [1] "5 148848523-148848559-004.jpeg"
## [1] "6 149013033-149013072-001.jpeg"
## [1] "7 149014631-149014646-001.jpeg"
## [1] "8 149014631-149014646-002.jpeg"
## [1] "9 149014631-149014646-003.jpeg"
## [1] "10 149014631-149014646-004.jpeg"
## [1] "11 149014631-149014654-001.jpeg"
## [1] "12 149014631-149014654-002.jpeg"
## [1] "13 149014631-149014654-003.jpeg"
## [1] "14 149014929-149014984-001.jpeg"
## [1] "15 149014929-149014994-001.jpeg"
## [1] "16 149014929-149015008-001.jpeg"
## [1] "17 149014929-149015008-002.jpeg"
## [1] "18 149014929-149015008-003.jpeg"
## [1] "19 149015814-149015831-001.jpeg"
## [1] "20 149053341-149053350-001.jpeg"
## [1] "21 149053341-149053350-002.jpeg"
## [1] "22 149053533-149053553-001.jpeg"
## [1] "23 149053945-149053969-001.jpeg"
## [1] "24 149053945-149053976-001.jpeg"
## [1] "25 149054169-149054199-001.jpeg"
## [1] "26 149054642-149054724-001.jpeg"
## [1] "27 149057261-149057272-001.jpeg"
## [1] "28 149057261-149057272-002.jpeg"
## [1] "29 149057756-149057766-001.jpeg"
## [1] "30 149057756-149057766-002.jpeg"
## [1] "31 149057756-149057776-001.jpeg"
## [1] "32 149057756-149057788-001.jpeg"
## [1] "33 149057756-149057788-002.jpeg"
## [1] "34 149057812-149057825-001.jpeg"
## [1] "35 149057812-149057825-003.jpeg"
## [1] "36 149057812-149057825-004.jpeg"
## [1] "37 149057812-149057825-005.jpeg"
## [1] "38 149057812-149057847-002.jpeg"
## [1] "39 149057812-149057847-003.jpeg"
## [1] "40 149057812-149057847-004.jpeg"
## [1] "41 149057812-149057847-005.jpeg"
## [1] "42 149057812-149057847-006.jpeg"
## [1] "43 149058262-149058272-003.jpeg"
## [1] "44 149058262-149058288-001.jpeg"
## [1] "45 149058262-149058288-002.jpeg"
## [1] "46 149058262-149058288-003.jpeg"
## [1] "47 149058262-149058288-004.jpeg"
## [1] "48 149058262-149058288-005.jpeg"
## [1] "49 149060626-149060638-001.jpeg"
## [1] "50 149060626-149060638-002.jpeg"
## [1] "51 149060932-149060940-001.jpeg"
## [1] "52 149060932-149060940-002.jpeg"
## [1] "53 149061666-149061683-003.jpeg"
## [1] "54 149061666-149061699-001.jpeg"
## [1] "55 149061666-149061699-003.jpeg"
## [1] "56 149062277-149062289-001.jpeg"
## [1] "57 149062593-149062608-001.jpeg"
## [1] "58 149096505-149096567-001.jpeg"
## [1] "59 149096505-149096601-001.jpeg"
## [1] "60 149097324-149097352-001.jpeg"
## [1] "61 149097324-149097373-001.jpeg"
## [1] "62 149097324-149097388-001.jpeg"
## [1] "63 149097594-149097637-001.jpeg"
## [1] "64 149097594-149097637-002.jpeg"
## [1] "65 149097594-149097661-001.jpeg"
## [1] "66 149098972-149099015-001.jpeg"
## [1] "67 149098972-149099015-002.jpeg"
## [1] "68 149099502-149099544-001.jpeg"
## [1] "69 149099680-149099714-001.jpeg"
## [1] "70 149100064-149100072-001.jpeg"
## [1] "71 149101437-149101456-001.jpeg"
## [1] "72 149101437-149101467-001.jpeg"
## [1] "73 149101437-149101467-002.jpeg"
## [1] "74 149101620-149101650-001.jpeg"
## [1] "75 149101620-149101650-002.jpeg"
## [1] "76 149101620-149101650-003.jpeg"
## [1] "77 149101620-149101650-004.jpeg"
## [1] "78 149101620-149101664-001.jpeg"
## [1] "79 149102510-149102519-001.jpeg"
## [1] "80 149102510-149102519-002.jpeg"
## [1] "81 149102510-149102527-001.jpeg"
## [1] "82 149102510-149102527-002.jpeg"
## [1] "83 149102510-149102527-003.jpeg"
## [1] "84 149105131-149105205-001.jpeg"
## [1] "85 149105131-149105205-002.jpeg"
## [1] "86 149105131-149105205-003.jpeg"
## [1] "87 149105131-149105205-004.jpeg"
## [1] "88 149105131-149105260-001.jpeg"
## [1] "89 149105131-149105260-002.jpeg"
## [1] "90 149105289-149105300-001.jpeg"
## [1] "91 149105289-149105300-002.jpeg"
## [1] "92 149105289-149105311-001.jpeg"
## [1] "93 149105289-149105311-002.jpeg"
## [1] "94 149105289-149105311-003.jpeg"
## [1] "95 149105289-149105311-004.jpeg"
## [1] "96 149140081-149140096-001.jpeg"
## [1] "97 149140081-149140096-002.jpeg"
## [1] "98 149140536-149140551-001.jpeg"
## [1] "99 149140849-149140856-001.jpeg"
## [1] "100 149143434-149143442-001.jpeg"
## [1] "101 149190861-149190868-001.jpeg"
## [1] "102 149315960-149315967-001.jpeg"
## [1] "103 149316754-149316762-002.jpeg"
## [1] "104 149316754-149316779-001.jpeg"
## [1] "105 149316754-149316779-002.jpeg"
## [1] "106 149316941-149316952-001.jpeg"
## [1] "107 149316941-149316961-002.jpeg"
## [1] "108 149357849-149357857-001.jpeg"
## [1] "109 149357849-149357857-002.jpeg"
## [1] "110 149357849-149357857-003.jpeg"
## [1] "111 149357849-149357867-001.jpeg"
## [1] "112 149357849-149357867-002.jpeg"
## [1] "113 149358108-149358125-001.jpeg"
## [1] "114 153275893-153275939-001.jpeg"
## [1] "115 153275893-153275939-003.jpeg"
## [1] "116 153276386-153276405-001.jpeg"
## [1] "117 153276386-153276405-002.jpeg"
## [1] "118 153276386-153276405-003.jpeg"
## [1] "119 153276386-153276405-004.jpeg"
## [1] "120 153276386-153276414-001.jpeg"
## [1] "121 153276799-153276818-002.jpeg"
## [1] "122 153697342-153697353-001.jpeg"
## [1] "123 153697342-153697366-001.jpeg"
## [1] "124 153697342-153697366-002.jpeg"
## [1] "125 153697342-153697375-001.jpeg"
## [1] "126 153697342-153697375-002.jpeg"
## [1] "127 153697342-153697375-003.jpeg"
## [1] "128 153697342-153697384-001.jpeg"
## [1] "129 153697342-153697384-002.jpeg"
## [1] "130 153698259-153698268-001.jpeg"
## [1] "131 153698259-153698268-002.jpeg"
## [1] "132 153698259-153698268-003.jpeg"
## [1] "133 153701139-153701158-001.jpeg"
## [1] "134 153701139-153701158-003.jpeg"
## [1] "135 153701139-153701158-004.jpeg"
## [1] "136 153701139-153701158-005.jpeg"
## [1] "137 153701622-153701650-001.jpeg"
## [1] "138 153829664-153829678-001.jpeg"
## [1] "139 153829700-153829711-001.jpeg"
## [1] "140 153829745-153829763-001.jpeg"
## [1] "141 153915230-153915252-001.jpeg"
## [1] "142 153916675-153916684-001.jpeg"
## [1] "143 204870815-204870836-001.jpeg"
## [1] "144 204870815-204870836-002.jpeg"
## [1] "145 204870892-204870902-001.jpeg"
## [1] "146 204870892-204870910-001.jpeg"
## [1] "1 153956040-153956058-001.jpeg"
## [1] "2 153956040-153956058-002.jpeg"
## [1] "3 153956040-153956058-003.jpeg"
## [1] "4 153956040-153956058-004.jpeg"
## [1] "5 153956040-153956058-005.jpeg"
## [1] "6 153956040-153956058-006.jpeg"
## [1] "7 153956040-153956072-001.jpeg"
## [1] "8 153956040-153956072-002.jpeg"
## [1] "9 153956040-153956072-003.jpeg"
## [1] "10 153956040-153956072-004.jpeg"
## [1] "11 153956444-153956458-001.jpeg"
## [1] "12 153956444-153956458-002.jpeg"
## [1] "13 153956444-153956458-003.jpeg"
## [1] "14 153956444-153956458-004.jpeg"
## [1] "15 153958154-153958168-001.jpeg"
## [1] "16 153958154-153958168-003.jpeg"
## [1] "17 153958154-153958179-001.jpeg"
## [1] "18 153958154-153958179-002.jpeg"
## [1] "19 153958154-153958194-001.jpeg"
## [1] "20 153958154-153958194-002.jpeg"
## [1] "21 153958154-153958194-003.jpeg"
## [1] "22 153958154-153958194-004.jpeg"
## [1] "23 153958154-153958211-001.jpeg"
## [1] "24 153958154-153958248-002.jpeg"
## [1] "25 153958154-153958248-003.jpeg"
## [1] "26 153958345-153958364-001.jpeg"
## [1] "27 153958547-153958572-001.jpeg"
## [1] "28 153958547-153958572-002.jpeg"
## [1] "29 153958547-153958572-003.jpeg"
## [1] "30 153958547-153958572-004.jpeg"
## [1] "31 153958547-153958572-005.jpeg"
## [1] "32 153958547-153958572-006.jpeg"
## [1] "33 153958547-153958572-007.jpeg"
## [1] "34 153958547-153958572-008.jpeg"
## [1] "35 153958547-153958572-009.jpeg"
## [1] "36 153958547-153958590-001.jpeg"
## [1] "37 153958547-153958590-002.jpeg"
## [1] "38 157181146-157181197-001.jpeg"
## [1] "39 157181146-157181197-002.jpeg"
## [1] "40 157181146-157181197-003.jpeg"
## [1] "41 157181481-157181497-001.jpeg"
## [1] "42 157181481-157181497-002.jpeg"
## [1] "43 157181481-157181497-003.jpeg"
## [1] "44 157181481-157181497-004.jpeg"
## [1] "45 157184850-157184920-001.jpeg"
## [1] "46 157184850-157184920-002.jpeg"
## [1] "47 157185433-157185454-001.jpeg"
## [1] "48 157185433-157185479-001.jpeg"
## [1] "49 157185433-157185479-002.jpeg"
## [1] "50 157185433-157185479-003.jpeg"
## [1] "51 157185433-157185479-004.jpeg"
## [1] "52 157185433-157185479-005.jpeg"
## [1] "53 157185433-157185479-006.jpeg"
## [1] "54 157185433-157185479-007.jpeg"
## [1] "55 157185433-157185508-001.jpeg"
## [1] "56 157185433-157185508-002.jpeg"
## [1] "57 157185433-157185508-003.jpeg"
## [1] "58 157185433-157185508-004.jpeg"
## [1] "59 157185433-157185508-005.jpeg"
## [1] "60 157185433-157185508-006.jpeg"
## [1] "61 157185527-157185540-001.jpeg"
## [1] "62 157185527-157185540-002.jpeg"
## [1] "63 157185527-157185540-003.jpeg"
## [1] "64 157185527-157185540-004.jpeg"
## [1] "65 157185730-157185758-001.jpeg"
## [1] "66 157185730-157185758-002.jpeg"
## [1] "67 157185781-157185793-001.jpeg"
## [1] "68 157185781-157185793-002.jpeg"
## [1] "69 157185781-157185793-003.jpeg"
## [1] "70 157185781-157185814-001.jpeg"
## [1] "71 157185781-157185814-002.jpeg"
## [1] "72 157224172-157224207-001.jpeg"
## [1] "73 157224172-157224207-002.jpeg"
## [1] "74 157224172-157224207-003.jpeg"
## [1] "75 157227058-157227087-001.jpeg"
## [1] "76 157227461-157227503-001.jpeg"
## [1] "77 157227461-157227503-002.jpeg"
## [1] "78 157227461-157227503-003.jpeg"
## [1] "79 157266930-157266947-001.jpeg"
## [1] "80 157266930-157266947-002.jpeg"
## [1] "81 157266930-157266947-003.jpeg"
## [1] "82 157267001-157267013-001.jpeg"
## [1] "83 157267059-157267072-001.jpeg"
## [1] "84 157267059-157267072-002.jpeg"
## [1] "85 157267059-157267072-003.jpeg"
## [1] "86 157267059-157267072-004.jpeg"
## [1] "87 157267263-157267286-001.jpeg"
## [1] "88 157267263-157267286-002.jpeg"
## [1] "89 158986766-158986776-001.jpeg"
## [1] "90 158986766-158986776-002.jpeg"
## [1] "91 158986813-158986820-001.jpeg"
## [1] "92 158986813-158986820-002.jpeg"
## [1] "93 158986920-158986928-001.jpeg"
## [1] "94 158986920-158986928-002.jpeg"
## [1] "95 158986920-158986928-003.jpeg"
## [1] "96 158986920-158986928-004.jpeg"
## [1] "97 158986920-158986928-005.jpeg"
## [1] "98 158986920-158986928-006.jpeg"
## [1] "1 153955676-153955721-001.jpeg"
## [1] "2 153955676-153955746-001.jpeg"
## [1] "3 153956279-153956296-001.jpeg"
## [1] "4 153958547-153958621-001.jpeg"
## [1] "5 157183332-157183388-001.jpeg"
## [1] "6 157183332-157183388-002.jpeg"
## [1] "7 157183412-157183656-001.jpeg"
## [1] "8 157183828-157183840-001.jpeg"
## [1] "9 157183828-157183877-001.jpeg"
## [1] "10 157183828-157183887-001.jpeg"
## [1] "11 157184013-157184023-001.jpeg"
## [1] "12 157184013-157184039-001.jpeg"
## [1] "13 157184067-157184080-001.jpeg"
## [1] "14 157184112-157184120-001.jpeg"
## [1] "15 157185575-157185586-001.jpeg"
## [1] "16 157185677-157185690-001.jpeg"
## [1] "17 157222534-157222561-001.jpeg"
## [1] "18 157222534-157222579-001.jpeg"
## [1] "19 157222534-157223349-001.jpeg"
## [1] "20 157222534-157223355-001.jpeg"
## [1] "21 157222534-157223364-001.jpeg"
## [1] "22 157222647-157222660-001.jpeg"
## [1] "23 157222647-157222687-001.jpeg"
## [1] "24 157222801-157222811-001.jpeg"
## [1] "25 157222801-157222811-002.jpeg"
## [1] "26 157222801-157222844-001.jpeg"
## [1] "27 157223246-157223258-001.jpeg"
## [1] "28 157223321-157223328-001.jpeg"
## [1] "29 157223321-157223341-001.jpeg"
## [1] "30 157223394-157223406-001.jpeg"
## [1] "31 157223394-157223406-002.jpeg"
## [1] "32 157223394-157223428-001.jpeg"
## [1] "33 157223394-157223428-002.jpeg"
## [1] "34 157223549-157223570-001.jpeg"
## [1] "35 157223779-157224140-001.jpeg"
## [1] "36 157224458-157224464-001.jpeg"
## [1] "37 157268504-157268534-001.jpeg"
## [1] "38 158989297-158989324-001.jpeg"
## [1] "39 209047526-209047626-001.jpeg"
## [1] "40 209047526-209047762-001.jpeg"
## [1] "41 209047881-209047928-001.jpeg"
## [1] "42 209048086-209048239-001.jpeg"
## [1] "43 209307421-209307504-001.jpeg"
## [1] "44 209307421-209307561-001.jpeg"
## [1] "45 209522316-209522391-001.jpeg"
## [1] "46 209522474-209522522-001.jpeg"
## [1] "47 209522474-209522554-001.jpeg"
## [1] "48 209522641-209522674-001.jpeg"
## [1] "49 209522800-209522835-001.jpeg"
## [1] "50 209522940-209522970-001.jpeg"
## [1] "51 209522940-209522991-001.jpeg"
## [1] "52 209522940-209523052-001.jpeg"
## [1] "53 209565409-209565466-001.jpeg"
## [1] "54 209565409-209565503-001.jpeg"
## [1] "55 209565409-209565600-001.jpeg"
## [1] "56 209565698-209565729-001.jpeg"
## [1] "57 209565698-209565772-001.jpeg"
## [1] "58 209565864-209565890-001.jpeg"
## [1] "59 209565864-209565911-001.jpeg"
## [1] "60 209565864-209565950-001.jpeg"
## [1] "61 209566047-209566095-001.jpeg"
## [1] "62 209566047-209566125-001.jpeg"
## [1] "63 209566205-209566247-001.jpeg"
## [1] "64 209566205-209566266-001.jpeg"
## [1] "65 209566205-209566289-001.jpeg"
## [1] "66 209566205-209566321-001.jpeg"
## [1] "67 209566205-209566333-001.jpeg"
## [1] "68 209566399-209566464-001.jpeg"
## [1] "69 209566399-209566485-001.jpeg"
## [1] "70 209566399-209566517-001.jpeg"
## [1] "1 153958345-153958392-001.jpeg"
## [1] "2 153960256-153960295-002.jpeg"
## [1] "3 157181281-157181308-001.jpeg"
## [1] "4 157181387-157181418-001.jpeg"
## [1] "5 157181525-157181540-001.jpeg"
## [1] "6 157181569-157181599-001.jpeg"
## [1] "7 157181671-157181686-001.jpeg"
## [1] "8 157181671-157181697-001.jpeg"
## [1] "9 157181671-157181708-001.jpeg"
## [1] "10 157183193-157183241-001.jpeg"
## [1] "11 157183193-157183279-001.jpeg"
## [1] "12 157183332-157183346-001.jpeg"
## [1] "13 157183332-157183361-001.jpeg"
## [1] "14 157183332-157183376-001.jpeg"
## [1] "15 157183412-157183614-001.jpeg"
## [1] "16 157183412-157183638-001.jpeg"
## [1] "17 157183722-157183783-001.jpeg"
## [1] "18 157183722-157183783-002.jpeg"
## [1] "19 157183964-157183980-001.jpeg"
## [1] "20 157183964-157183991-001.jpeg"
## [1] "21 157184013-157184031-001.jpeg"
## [1] "22 157184112-157184128-001.jpeg"
## [1] "23 157184850-157184870-001.jpeg"
## [1] "24 157184850-157184891-001.jpeg"
## [1] "25 157184850-157184942-001.jpeg"
## [1] "26 157184850-157184976-001.jpeg"
## [1] "27 157184999-157185032-001.jpeg"
## [1] "28 157185054-157185079-001.jpeg"
## [1] "29 157185133-157185143-001.jpeg"
## [1] "30 157185164-157185177-001.jpeg"
## [1] "31 157185164-157185196-001.jpeg"
## [1] "32 157185208-157185238-001.jpeg"
## [1] "33 157185309-157185360-001.jpeg"
## [1] "34 157185613-157185627-001.jpeg"
## [1] "35 157222534-157222607-001.jpeg"
## [1] "36 157222737-157222750-001.jpeg"
## [1] "37 157222737-157222750-002.jpeg"
## [1] "38 157222904-157223196-001.jpeg"
## [1] "39 157222904-157223217-001.jpeg"
## [1] "40 157223659-157223667-001.jpeg"
## [1] "41 157223659-157223667-002.jpeg"
## [1] "42 157223735-157223751-001.jpeg"
## [1] "43 157223735-157223766-001.jpeg"
## [1] "44 157223779-157224126-001.jpeg"
## [1] "45 157224172-157224179-001.jpeg"
## [1] "46 157224244-157224258-001.jpeg"
## [1] "47 157224297-157224304-001.jpeg"
## [1] "48 157224297-157224304-002.jpeg"
## [1] "49 157224297-157224320-001.jpeg"
## [1] "50 157224297-157224320-002.jpeg"
## [1] "51 157224297-157224336-001.jpeg"
## [1] "52 157224352-157224395-001.jpeg"
## [1] "53 157224412-157224429-001.jpeg"
## [1] "54 157224458-157224483-001.jpeg"
## [1] "55 157224504-157224520-001.jpeg"
## [1] "56 157267647-157267732-001.jpeg"
## [1] "57 157268242-157268296-001.jpeg"
## [1] "58 157268342-157268376-001.jpeg"
## [1] "59 157268342-157268401-001.jpeg"
## [1] "60 157268504-157268544-001.jpeg"
## [1] "61 157268587-157268617-001.jpeg"
## [1] "62 158987033-158987057-001.jpeg"
## [1] "63 158987453-158987462-001.jpeg"
## [1] "64 158987493-158987499-001.jpeg"
## [1] "65 158987493-158987505-001.jpeg"
## [1] "66 209047342-209047400-001.jpeg"
## [1] "67 209047342-209047443-001.jpeg"
## [1] "68 209047342-209047478-001.jpeg"
## [1] "69 209047526-209047717-001.jpeg"
## [1] "70 209047526-209047798-001.jpeg"
## [1] "71 209047881-209048017-001.jpeg"
## [1] "72 209048086-209048137-001.jpeg"
## [1] "73 209048086-209048278-001.jpeg"
## [1] "74 209307421-209307597-001.jpeg"
## [1] "1 149056321-149056343-001.jpeg"
## [1] "2 149056321-149056343-002.jpeg"
## [1] "3 149056321-149056343-003.jpeg"
## [1] "4 149056321-149056360-001.jpeg"
## [1] "5 149056321-149056360-002.jpeg"
## [1] "6 149056321-149056360-003.jpeg"
## [1] "7 149056410-149056423-001.jpeg"
## [1] "8 149056410-149056423-002.jpeg"
## [1] "9 149056410-149056423-003.jpeg"
## [1] "10 149056410-149056423-004.jpeg"
## [1] "11 149056410-149056444-003.jpeg"
## [1] "12 149056410-149056444-004.jpeg"
## [1] "13 149056410-149056444-005.jpeg"
## [1] "14 149056410-149056458-001.jpeg"
## [1] "15 149056410-149056458-002.jpeg"
## [1] "16 149056410-149056458-004.jpeg"
## [1] "17 149056410-149056458-005.jpeg"
## [1] "18 149056545-149056555-001.jpeg"
## [1] "19 149056545-149056555-002.jpeg"
## [1] "20 149056785-149056797-001.jpeg"
## [1] "21 149056785-149056797-003.jpeg"
## [1] "22 149056785-149056797-004.jpeg"
## [1] "23 149056785-149056813-001.jpeg"
## [1] "24 149056785-149056813-002.jpeg"
## [1] "25 149056785-149056813-004.jpeg"
## [1] "26 149056785-149056813-006.jpeg"
## [1] "27 149056785-149056813-007.jpeg"
## [1] "28 149056873-149056884-001.jpeg"
## [1] "29 149056873-149056884-002.jpeg"
## [1] "30 149056933-149056945-002.jpeg"
## [1] "31 149056933-149056945-004.jpeg"
## [1] "32 149056933-149056945-005.jpeg"
## [1] "33 149056933-149056982-002.jpeg"
## [1] "34 149056933-149056982-003.jpeg"
## [1] "35 149056933-149056982-005.jpeg"
## [1] "36 149056933-149056982-006.jpeg"
## [1] "37 149057164-149057176-001.jpeg"
## [1] "38 149057164-149057197-001.jpeg"
## [1] "39 149057164-149057197-002.jpeg"
## [1] "40 149058170-149058202-001.jpeg"
## [1] "41 149058170-149058202-002.jpeg"
## [1] "42 149058170-149058218-001.jpeg"
## [1] "43 149058170-149058218-002.jpeg"
## [1] "44 149058170-149058218-003.jpeg"
## [1] "45 149058170-149058218-004.jpeg"
## [1] "46 149058262-149058309-001.jpeg"
## [1] "47 149061221-149061261-002.jpeg"
## [1] "48 149061221-149061289-001.jpeg"
## [1] "49 149061548-149061568-002.jpeg"
## [1] "50 149061548-149061568-003.jpeg"
## [1] "51 149061548-149061568-004.jpeg"
## [1] "52 149062753-149062761-001.jpeg"
## [1] "53 149096505-149096614-001.jpeg"
## [1] "54 149096505-149096614-002.jpeg"
## [1] "55 149096505-149096614-003.jpeg"
## [1] "56 149096854-149096870-001.jpeg"
## [1] "57 149096927-149096943-001.jpeg"
## [1] "58 149096927-149096943-002.jpeg"
## [1] "59 149098972-149098988-001.jpeg"
## [1] "60 149098972-149098988-002.jpeg"
## [1] "61 149099096-149099104-001.jpeg"
## [1] "62 149100006-149100028-001.jpeg"
## [1] "63 149100006-149100028-002.jpeg"
## [1] "64 149101894-149101909-001.jpeg"
## [1] "65 149101894-149101909-002.jpeg"
## [1] "66 149101894-149101909-003.jpeg"
## [1] "67 149101894-149101918-001.jpeg"
## [1] "68 149101894-149101930-001.jpeg"
## [1] "69 149101894-149101930-002.jpeg"
## [1] "70 149101894-149101930-003.jpeg"
## [1] "71 149105131-149105155-001.jpeg"
## [1] "72 149105131-149105155-002.jpeg"
## [1] "73 149105131-149105155-003.jpeg"
## [1] "74 149105131-149105155-004.jpeg"
## [1] "75 149142457-149142465-001.jpeg"
## [1] "76 149143028-149143108-001.jpeg"
## [1] "77 149143028-149143108-002.jpeg"
## [1] "78 149143295-149143307-001.jpeg"
## [1] "79 149143469-149143479-002.jpeg"
## [1] "80 149143469-149143497-001.jpeg"
## [1] "81 149143469-149143497-002.jpeg"
## [1] "82 149143469-149143497-003.jpeg"
## [1] "83 149181806-149181819-001.jpeg"
## [1] "84 149181806-149181819-002.jpeg"
## [1] "85 149181806-149181819-003.jpeg"
## [1] "86 149183509-149183519-001.jpeg"
## [1] "87 149186957-149186968-001.jpeg"
## [1] "88 149186957-149186968-002.jpeg"
## [1] "89 149314587-149314605-001.jpeg"
## [1] "90 149314679-149314719-001.jpeg"
## [1] "91 149315229-149315291-001.jpeg"
## [1] "92 149315229-149315291-002.jpeg"
## [1] "93 149315229-149315291-003.jpeg"
## [1] "94 149315229-149315310-001.jpeg"
## [1] "95 149315229-149315310-002.jpeg"
## [1] "96 149316754-149316795-002.jpeg"
## [1] "97 149317002-149317010-001.jpeg"
## [1] "98 149317002-149317010-002.jpeg"
## [1] "99 149317002-149317010-003.jpeg"
## [1] "100 149317204-149317230-001.jpeg"
## [1] "101 149317204-149317230-002.jpeg"
## [1] "102 149317204-149317230-003.jpeg"
## [1] "103 153314857-153314909-001.jpeg"
## [1] "104 153314857-153314909-002.jpeg"
## [1] "105 153314956-153314978-003.jpeg"
## [1] "106 153314956-153314985-001.jpeg"
## [1] "107 153314956-153314985-002.jpeg"
## [1] "108 153314956-153314996-001.jpeg"
## [1] "109 153315615-153315624-001.jpeg"
## [1] "110 153315615-153315624-002.jpeg"
## [1] "111 153315615-153315629-001.jpeg"
## [1] "112 153315615-153315629-003.jpeg"
## [1] "113 153654325-153654351-001.jpeg"
## [1] "114 153654325-153654351-002.jpeg"
## [1] "115 153654325-153654351-003.jpeg"
## [1] "116 153654325-153654351-004.jpeg"
## [1] "117 153654325-153654360-001.jpeg"
## [1] "118 153654325-153654360-002.jpeg"
## [1] "119 153654325-153654360-003.jpeg"
## [1] "120 153654869-153654880-001.jpeg"
## [1] "121 153654869-153654880-003.jpeg"
## [1] "122 153654869-153654880-004.jpeg"
## [1] "123 153655016-153655039-001.jpeg"
## [1] "124 153655016-153655039-003.jpeg"
## [1] "125 153655016-153655039-004.jpeg"
## [1] "126 153655016-153655039-005.jpeg"
## [1] "127 153655087-153655101-002.jpeg"
## [1] "128 153655087-153655101-003.jpeg"
## [1] "129 153657121-153657151-001.jpeg"
## [1] "130 153657121-153657160-001.jpeg"
## [1] "131 153657121-153657160-002.jpeg"
## [1] "132 153659229-153659243-001.jpeg"
## [1] "133 153659229-153659243-002.jpeg"
## [1] "134 153659229-153659243-003.jpeg"
## [1] "135 153659229-153659256-001.jpeg"
## [1] "136 153659229-153659256-002.jpeg"
## [1] "137 153659229-153659256-003.jpeg"
## [1] "138 153659229-153659264-001.jpeg"
## [1] "139 153826963-153826970-001.jpeg"
## [1] "140 153826963-153826976-001.jpeg"
## [1] "141 153826963-153826976-002.jpeg"
## [1] "142 153826963-153826976-003.jpeg"
## [1] "143 153826963-153826982-001.jpeg"
## [1] "144 153826963-153826982-002.jpeg"
## [1] "145 153826963-153826982-003.jpeg"
## [1] "146 153826963-153826989-001.jpeg"
## [1] "147 153826963-153826989-002.jpeg"
## [1] "148 153827595-153827604-001.jpeg"
## [1] "149 153827595-153827604-002.jpeg"
## [1] "150 153827595-153827604-004.jpeg"
## [1] "151 153827595-153827611-001.jpeg"
## [1] "152 153827595-153827611-002.jpeg"
## [1] "153 153827595-153827618-001.jpeg"
## [1] "154 153827595-153827618-002.jpeg"
## [1] "155 153828877-153828890-001.jpeg"
## [1] "156 153828877-153828890-002.jpeg"
## [1] "157 153828877-153828897-001.jpeg"
## [1] "158 153828877-153828902-002.jpeg"
## [1] "159 153828952-153828970-001.jpeg"
## [1] "160 153828952-153828980-001.jpeg"
## [1] "161 153828952-153829000-001.jpeg"
## [1] "162 153828952-153829000-002.jpeg"
## [1] "163 153828952-153829000-003.jpeg"
## [1] "164 153828952-153829005-001.jpeg"
## [1] "165 153828952-153829005-002.jpeg"
## [1] "166 153828952-153829009-002.jpeg"
## [1] "167 153829063-153829073-001.jpeg"
## [1] "168 153829063-153829073-002.jpeg"
## [1] "169 153829063-153829078-001.jpeg"
## [1] "170 153829063-153829086-001.jpeg"
## [1] "171 153829745-153829768-001.jpeg"
## [1] "172 153829745-153829768-002.jpeg"
## [1] "173 153829745-153829768-003.jpeg"
## [1] "174 153830261-153830340-001.jpeg"
## [1] "175 153830261-153830340-002.jpeg"
## [1] "176 153830680-153830827-001.jpeg"
## [1] "177 153830680-153830827-002.jpeg"
## [1] "178 153830680-153830827-003.jpeg"
## [1] "179 153831352-153831372-001.jpeg"
## [1] "180 153831352-153831372-002.jpeg"
## [1] "181 153831352-153831372-003.jpeg"
## [1] "182 153831352-153831372-004.jpeg"
## [1] "183 153915316-153915330-003.jpeg"
## [1] "184 153915432-153915441-002.jpeg"
## [1] "185 153915432-153915472-001.jpeg"
## [1] "186 153915634-153915665-001.jpeg"
## [1] "187 153915634-153915665-002.jpeg"
## [1] "188 153915634-153915665-003.jpeg"
## [1] "189 153915634-153915665-005.jpeg"
## [1] "190 153915634-153915665-006.jpeg"
## [1] "191 153915634-153915665-007.jpeg"
## [1] "192 153915726-153915736-001.jpeg"
## [1] "193 153915726-153915774-001.jpeg"
## [1] "194 153916001-153916020-001.jpeg"
## [1] "195 153916001-153916020-003.jpeg"
## [1] "196 153916114-153916129-002.jpeg"
## [1] "197 153916922-153916963-001.jpeg"

#ts <- sapply(listFiles, function(x) images2matrix(x, w=28, h=28, class= seq_len(length(listFiles)) ))

classes <- list(class1_mat, class2_mat, class3_mat) #, class4_mat, class5_mat, class6_mat, class7_mat)

class2_mat[1:4, 1:4]

#class2_mat[1:4, 1:4]
dim(class1_mat)

## [1]  70 785

#dim(class2_mat)

Visualize image directly from row

## visualize image from vector
vec2img <- function(df, nrow){
  
  ##  plot(EBImage::Image(x_test[1,], dim = c(28,28) ))
  
 i <- EBImage::Image(as.numeric(df[nrow,]))
sqr <- sqrt(length(df[nrow,])) 
dim(i) <- c(sqr,sqr, 1)
i <- EBImage::resize(i, w= 156, h= 156)
 
 return(plot(i))
}
vec2img(class1_mat[,-1], 65)

Sampling Training and Testing dataset

#Train1 <-  class1_mat[sample(nrow(class1_mat),size=60,replace=FALSE),]
#Train2 <- class2_mat[sample(nrow(class2_mat), size = 130, replace = FALSE),]

## Sampling training and testing row from dataframe
sampling_train_test <- function(df, n_test= 5 ){
  
  test <- dplyr::sample_n(df, n_test )
  
  ## omit test rows from train and remove duplicates if exist
  
  ## NOT RIGHT  NOT RIGHT NOT RIGHT NOT RIGHT NOT RIGHT 
   train <-  df[-duplicated(rbind(test, df)),]   
  
  ## correct sampling of train
  # train <- df[!(rownames(df) %in% rownames(test)),]
  
  return(list(train = train, test = test))
}

# function to merge training and testing dataset from a list of dataframes (classes)
merge_train_test <- function(list.df, n_test = 5){
  
  set.seed(1234)
  
  list_train_test <- lapply(list.df,sampling_train_test)
  
  sum_train <- data.frame()
  sum_test <- data.frame()
  for(i in 1:length(list_train_test)){
    
   sum_train <-rbind( sum_train, list_train_test[[i]]$train)
   
   sum_test <- rbind( sum_test, list_train_test[[i]]$test)
    
  }
  return(list(allTrain = sum_train, allTest = sum_test))
    
}

list_smp <- merge_train_test(classes, n_test= 5)

Train <- list_smp$allTrain
Test <- list_smp$allTest


Test[, 1:20]

The sampling code was done with `train <- df[-duplicated(rbind(test, df)),]`. It is recommandet to continue the code with `train <- df[!(rownames(df) %in% rownames(test)),]`

# Set up train and test arrays
train <- data.matrix(Train)
train_x <- t(train[, -1])
train_y <- train[, 1]
train_array <- train_x
dim(train_array) <- c(28, 28, 1, ncol(train_x))

test <- data.matrix(Test)
test_x <- t(test[, -1])
test_y <- test[, 1]
test_array <- test_x
dim(test_array) <- c(28, 28, 1, ncol(test_x))

# Set up the symbolic model
#-------------------------------------------------------------------------------
require(mxnet)

## Loading required package: mxnet

data <- mx.symbol.Variable('data')
# 1st convolutional layer
conv_1 <- mx.symbol.Convolution(data = data, kernel = c(5, 5), num_filter = 20)
tanh_1 <- mx.symbol.Activation(data = conv_1, act_type = "tanh")
pool_1 <- mx.symbol.Pooling(data = tanh_1, pool_type = "max", kernel = c(2, 2), stride = c(2, 2))
# 2nd convolutional layer
conv_2 <- mx.symbol.Convolution(data = pool_1, kernel = c(5, 5), num_filter = 50)
tanh_2 <- mx.symbol.Activation(data = conv_2, act_type = "tanh")
pool_2 <- mx.symbol.Pooling(data=tanh_2, pool_type = "max", kernel = c(2, 2), stride = c(2, 2))
# 1st fully connected layer
flatten <- mx.symbol.Flatten(data = pool_2)
fc_1 <- mx.symbol.FullyConnected(data = flatten, num_hidden = 500)
tanh_3 <- mx.symbol.Activation(data = fc_1, act_type = "tanh")
# 2nd fully connected layer
fc_2 <- mx.symbol.FullyConnected(data = tanh_3, num_hidden = 40)
# Output. Softmax output since we'd like to get some probabilities.
NN_model <- mx.symbol.SoftmaxOutput(data = fc_2)

require(magrittr)
# Pre-training set up
#-------------------------------------------------------------------------------

# Set seed for reproducibility
mx.set.seed(100)

# Device used. CPU in my case.
devices <- mx.cpu()

# Training
#-------------------------------------------------------------------------------

# Train the model
model <- mx.model.FeedForward.create(symbol = NN_model,       # The network schema
                                     X = train_array,         # Training array
                                     y = train_y,             # Labels/classes of training dataset
                                     ctx = devices,
                                     num.round = 150,
                                     array.batch.size = 20,  # number of array in the batch size
                                     learning.rate = 0.02,
                                     momentum = 0.9,
                                     optimizer = "sgd",
                                     eval.metric = mx.metric.accuracy,
                                     #initializer=mx.init.uniform(0.05),
                                     epoch.end.callback = mx.callback.log.train.metric(100))

## Start training with 1 devices

## [1] Train-accuracy=0.366666673372189

## [2] Train-accuracy=0.408333341280619

## [3] Train-accuracy=0.408333341280619

## [4] Train-accuracy=0.35416667163372

## [5] Train-accuracy=0.329166673123837

## [6] Train-accuracy=0.41250000645717

## [7] Train-accuracy=0.375000006208817

## [8] Train-accuracy=0.262500004842877

## [9] Train-accuracy=0.358333339293798

## [10] Train-accuracy=0.333333340783914

## [11] Train-accuracy=0.316666672627131

## [12] Train-accuracy=0.387500005463759

## [13] Train-accuracy=0.287500004594525

## [14] Train-accuracy=0.408333341280619

## [15] Train-accuracy=0.395833340783914

## [16] Train-accuracy=0.325000004221996

## [17] Train-accuracy=0.408333338797092

## [18] Train-accuracy=0.395833338300387

## [19] Train-accuracy=0.391666673123837

## [20] Train-accuracy=0.395833340783914

## [21] Train-accuracy=0.387500005463759

## [22] Train-accuracy=0.387500005463759

## [23] Train-accuracy=0.387500005463759

## [24] Train-accuracy=0.387500005463759

## [25] Train-accuracy=0.387500005463759

## [26] Train-accuracy=0.387500005463759

## [27] Train-accuracy=0.39166667064031

## [28] Train-accuracy=0.395833338300387

## [29] Train-accuracy=0.387500005463759

## [30] Train-accuracy=0.387500002980232

## [31] Train-accuracy=0.375000002483527

## [32] Train-accuracy=0.379166670143604

## [33] Train-accuracy=0.387500002980232

## [34] Train-accuracy=0.387500002980232

## [35] Train-accuracy=0.387500002980232

## [36] Train-accuracy=0.379166670143604

## [37] Train-accuracy=0.370833337306976

## [38] Train-accuracy=0.370833334823449

## [39] Train-accuracy=0.387500005463759

## [40] Train-accuracy=0.383333340287209

## [41] Train-accuracy=0.408333336313566

## [42] Train-accuracy=0.400000003476938

## [43] Train-accuracy=0.433333339790503

## [44] Train-accuracy=0.487500004470348

## [45] Train-accuracy=0.550000006953875

## [46] Train-accuracy=0.516666665673256

## [47] Train-accuracy=0.550000006953875

## [48] Train-accuracy=0.545833339293798

## [49] Train-accuracy=0.570833340287209

## [50] Train-accuracy=0.549999992052714

## [51] Train-accuracy=0.620833337306976

## [52] Train-accuracy=0.620833332339923

## [53] Train-accuracy=0.641666655739148

## [54] Train-accuracy=0.675000001986822

## [55] Train-accuracy=0.712500000993411

## [56] Train-accuracy=0.704166668156783

## [57] Train-accuracy=0.720833336313566

## [58] Train-accuracy=0.750000004967054

## [59] Train-accuracy=0.737500006953875

## [60] Train-accuracy=0.75

## [61] Train-accuracy=0.762499998013178

## [62] Train-accuracy=0.72916667163372

## [63] Train-accuracy=0.716666673620542

## [64] Train-accuracy=0.662499999006589

## [65] Train-accuracy=0.562500002483527

## [66] Train-accuracy=0.662500001490116

## [67] Train-accuracy=0.75

## [68] Train-accuracy=0.729166666666667

## [69] Train-accuracy=0.745833337306976

## [70] Train-accuracy=0.745833337306976

## [71] Train-accuracy=0.791666666666667

## [72] Train-accuracy=0.79166667163372

## [73] Train-accuracy=0.812500009934107

## [74] Train-accuracy=0.841666663686434

## [75] Train-accuracy=0.862499992052714

## [76] Train-accuracy=0.854166666666667

## [77] Train-accuracy=0.849999994039536

## [78] Train-accuracy=0.866666664679845

## [79] Train-accuracy=0.895833333333333

## [80] Train-accuracy=0.862500001986822

## [81] Train-accuracy=0.866666664679845

## [82] Train-accuracy=0.904166658719381

## [83] Train-accuracy=0.904166658719381

## [84] Train-accuracy=0.841666663686434

## [85] Train-accuracy=0.866666664679845

## [86] Train-accuracy=0.820833335320155

## [87] Train-accuracy=0.874999995032946

## [88] Train-accuracy=0.899999991059303

## [89] Train-accuracy=0.925000001986822

## [90] Train-accuracy=0.95833332836628

## [91] Train-accuracy=0.94166665772597

## [92] Train-accuracy=0.870833327372869

## [93] Train-accuracy=0.89166667064031

## [94] Train-accuracy=0.916666661699613

## [95] Train-accuracy=0.962499991059303

## [96] Train-accuracy=0.954166660706202

## [97] Train-accuracy=0.933333327372869

## [98] Train-accuracy=0.958333323399226

## [99] Train-accuracy=0.962499996026357

## [100] Train-accuracy=0.912499999006589

## [101] Train-accuracy=0.891666655739148

## [102] Train-accuracy=0.866666659712791

## [103] Train-accuracy=0.891666665673256

## [104] Train-accuracy=0.949999993046125

## [105] Train-accuracy=0.962499996026357

## [106] Train-accuracy=0.979166661699613

## [107] Train-accuracy=0.995833332339923

## [108] Train-accuracy=0.991666664679845

## [109] Train-accuracy=0.966666663686434

## [110] Train-accuracy=0.98333332935969

## [111] Train-accuracy=0.991666664679845

## [112] Train-accuracy=1

## [113] Train-accuracy=1

## [114] Train-accuracy=1

## [115] Train-accuracy=1

## [116] Train-accuracy=1

## [117] Train-accuracy=1

## [118] Train-accuracy=1

## [119] Train-accuracy=1

## [120] Train-accuracy=1

## [121] Train-accuracy=1

## [122] Train-accuracy=1

## [123] Train-accuracy=1

## [124] Train-accuracy=1

## [125] Train-accuracy=1

## [126] Train-accuracy=1

## [127] Train-accuracy=1

## [128] Train-accuracy=1

## [129] Train-accuracy=1

## [130] Train-accuracy=1

## [131] Train-accuracy=1

## [132] Train-accuracy=1

## [133] Train-accuracy=1

## [134] Train-accuracy=1

## [135] Train-accuracy=1

## [136] Train-accuracy=1

## [137] Train-accuracy=1

## [138] Train-accuracy=1

## [139] Train-accuracy=1

## [140] Train-accuracy=1

## [141] Train-accuracy=1

## [142] Train-accuracy=1

## [143] Train-accuracy=1

## [144] Train-accuracy=1

## [145] Train-accuracy=1

## [146] Train-accuracy=1

## [147] Train-accuracy=1

## [148] Train-accuracy=1

## [149] Train-accuracy=1

## [150] Train-accuracy=1

summarymxnet <- summary(model$arg.params)
 data.frame(do.call(rbind, list(summarymxnet))) %>% 
  tibble::rownames_to_column("layers")

#Predict labels
predicted <- predict(model, test_array)
predicted <- mxnet:::predict.MXFeedForwardModel(model = model, X = test_array)
# Assign labels
predicted_labels <- max.col(t(predicted)) -1
# Get accuracy
table(test_y, predicted_labels)

##       predicted_labels
## test_y 1 2 3
##      1 5 0 0
##      2 0 5 0
##      3 0 0 5

predicted[1:8, 1:7]

##              [,1]         [,2]         [,3]         [,4]         [,5]
## [1,] 1.840857e-07 1.043327e-07 2.483293e-08 1.321998e-10 7.084486e-10
## [2,] 9.982890e-01 9.996439e-01 9.998586e-01 9.999977e-01 1.000000e+00
## [3,] 3.513488e-04 2.989487e-04 1.398128e-04 5.564680e-12 4.687845e-08
## [4,] 1.352182e-03 5.273091e-05 1.591530e-06 2.290906e-06 1.160615e-09
## [5,] 1.856780e-07 1.060495e-07 2.452958e-08 1.375781e-10 6.894709e-10
## [6,] 1.830753e-07 1.045981e-07 2.570320e-08 1.345270e-10 7.358344e-10
## [7,] 1.815529e-07 1.025818e-07 2.358473e-08 1.390918e-10 6.840624e-10
## [8,] 1.825740e-07 1.006227e-07 2.405853e-08 1.265014e-10 6.660885e-10
##              [,6]         [,7]
## [1,] 9.981917e-08 1.492411e-06
## [2,] 1.135575e-06 7.383204e-05
## [3,] 9.999942e-01 9.998740e-01
## [4,] 4.002962e-07 5.377533e-08
## [5,] 9.318004e-08 1.412985e-06
## [6,] 1.016199e-07 1.570883e-06
## [7,] 9.241506e-08 1.329237e-06
## [8,] 1.020780e-07 1.466257e-06

## get means of identical labels, located in the diagonal
sum(diag(table(test_y,predicted_labels)))/length(test_y)

## [1] 1

how can plot image from selected test_y Labels/classes

## The colnames of test_y in the rownames of the vector image
test_y

##  8 43 42 68 57 63  1 23 64 49 52 40 21 66 72 
##  1  1  1  1  1  2  2  2  2  2  3  3  3  3  3

test[1:5, 1:5]

##    Labels   pixel 1   pixel 2   pixel 3   pixel 4
## 8       1 0.5931481 0.7036114 0.6982935 0.7026261
## 43      1 0.6423086 0.6377714 0.6519608 0.6413866
## 42      1 0.6633787 0.6698980 0.6744198 0.6702614
## 68      1 0.6260237 0.6161531 0.5874750 0.5748833
## 57      1 0.2169985 0.1722289 0.2117230 0.2125759

## we would like to select classes 3 
selected_vec <- names(test_y[test_y==3])
#class(test[selected_vec,][,-1][1:5, 1:6])

for(i in 1: length(selected_vec)){
vec2img(Test[selected_vec,][,-1],i)
  
}

mxnet example 3 : PapSmear 2005 (Convoluted Neural Networks - CNN)