Create a loop to transform several images to vectors and join them into a dataframe

[mariagarza]

Hello everybody,

I have to develop a script in R that saves the vectors resulting from the transformation of several images in a data frame that will have as many rows as images (1000) and as many columns as variables (4096). The data frame will be exported to a csv file.

I have tried a loop so that it reads the images and transforms them one by one and then join them in a matrix or data frame. I'm new to this... I am lost... I share you what I have tried. Any idea or improvement?

set.seed(1234)
mypath = "C:/dataset/dataset/effusion/" #file where the images are
files <- list.files(path=mypath, pattern=".png$")
length(files)

for (i in files){
 im <- readImage(mypath[i], header = FALSE)
 grises <- rgb_2gray(im[i])
 # Resize the image to a size of 64 x64 pixels:
 resiz = resizeImage(grises[i], width = 64, height = 64, method = 'nearest')
 # Convert to vector:
 im.convert <- as.vector(resiz[i])
 out <- as.data.frame(do.call(rbind, im.convert)) # create a data frame
     }
  return(out)

Thank you very much!

[technocrat]

Hi, and welcome!

Please see the FAQ: What's a reproducible example (`reprex`) and how do I do one? Using a reprex, complete with representative data will attract quicker and more answers.

Here, to test my own answer I'm flying blind without representative data and the libraries being used for the functions like rgb_2gray. And the friction of reverse engineering the problem deters many who might answer.

It doesn't have to be *all the data, just representative, say 4x40 or whatever is big enough to yield a 64*64 image (whether or not it's the whole image. I'm thinking that a purrr:map will be quicker and easier than a loop.

[mariagarza]

@technocrat

Hi, and welcome!

Please see the FAQ: What's a reproducible example (reprex) and how do I do one? Using a reprex, complete with representative data will attract quicker and more answers.

Here, to test my own answer I'm flying blind without representative data and the libraries being used for the functions like rgb_2gray . And the friction of reverse engineering the problem deters many who might answer.

It doesn't have to be * all the data, just representative, say 4x40 or whatever is big enough to yield a 64∗6464*64 image (whether or not it's the whole image. I'm thinking that a purrr:map will be quicker and easier than a loop.

Thank you! I will try to do it with the advice you have given me.

On the other hand, and for more information, the data I have is a folder with 500 images of 512 x 512 pixels and 3 channels in a numerical vector of 4096 values. I must transform each image to 64 x 64 and convert the matrix into a vector. Then join all the vectors of each image in a csv. which will have one row for each image and 4096 columns.
I will try to make it with purr.

Thanks!