Feature/flow images from dataframe

[dfalbel]

This will add flow_images_from_dataframe function. Fixing #658.
Needs more testing before merging.

[jjallaire]

If the function itself requires pandas we should note this in the documentation (and also do a check during the call to print a clear error message to that effect). Note that pandas will generally not be installed in TF environments installed via install_tensorflow() without special extra effort.

[dfalbel]

Thanks JJ! I added a note to the docs here: https://github.com/rstudio/keras/pull/725/files#diff-151c30ec9c406dc7d593b4eeca5113a3R890 and I'll catch the error to print a clear message here: https://github.com/rstudio/keras/pull/725/files#diff-151c30ec9c406dc7d593b4eeca5113a3R910

[jjallaire]

This does leave open whether they used virtualenv or condaenv to install (the code above will default to virtualenv).

They could also do this: install_tensorflow(extra_packages = "pandas"). Not sure if that's a better thing to document or not though (will leave that to your discretion!)

[dfalbel]

Ok, I thnk it's better now: 643bca3#diff-151c30ec9c406dc7d593b4eeca5113a3R914

[jjallaire]

Okay, that looks great!

…

On Wed, Apr 3, 2019 at 1:15 PM Daniel Falbel ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In R/preprocessing.R <#725 (comment)>: > +#' @section Yields: `(x, y)` where `x` is an array of image data and `y` is a +#' array of corresponding labels. The generator loops indefinitely. +#' +#' @family image preprocessing +#' @export +flow_images_from_dataframe <- function( + dataframe, directory = NULL, x_col = "filename", y_col = "class", + generator = image_data_generator(), target_size = c(256,256), + color_mode = "rgb", classes = NULL, class_mode = "categorical", + batch_size = 32, shuffle = TRUE, seed = NULL, save_to_dir = NULL, + save_prefix = "", save_format = "png", subset = NULL, + interpolation = "nearest", drop_duplicates = TRUE) { + + if (!reticulate::py_module_available("pandas")) + stop("Pandas (python module) must be installed in the same environment as Keras.", + '. Install it using reticulate::py_install("pandas", envname = "r-tensorflow").') Ok, I thnk it's better now: 643bca3 #diff-151c30ec9c406dc7d593b4eeca5113a3R914 <643bca3#diff-151c30ec9c406dc7d593b4eeca5113a3R914> — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#725 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAGXx3kb4FcutAGyerkI95ghGCNcioZfks5vdOGhgaJpZM4cY7ij> .