R/ocr.R

In ropensci/tesseract: Open Source OCR Engine

#' Tesseract OCR
#'
#' Extract text from an image. Requires that you have training data for the language you
#' are reading. Works best for images with high contrast, little noise and horizontal text.
#' See [tesseract wiki](https://github.com/tesseract-ocr/tesseract/wiki/ImproveQuality) and
#' our package vignette for image preprocessing tips.
#'
#' The `ocr()` function returns plain text by default, or hOCR text if hOCR is set to `TRUE`.
#' The `ocr_data()` function returns a data frame with a confidence rate and bounding box for
#' each word in the text.
#'
#' @export
#' @useDynLib tesseract
#' @family tesseract
#' @param image file path, url, or raw vector to image (png, tiff, jpeg, etc)
#' @param engine a tesseract engine created with [tesseract()]. Alternatively a
#' language string which will be passed to [tesseract()].
#' @param HOCR if `TRUE` return results as HOCR xml instead of plain text
#' @rdname ocr
#' @references [Tesseract: Improving Quality](https://github.com/tesseract-ocr/tesseract/wiki/ImproveQuality)
#' @importFrom Rcpp sourceCpp
#' @examples # Simple example
#' text <- ocr("https://jeroen.github.io/images/testocr.png")
#' cat(text)
#'
#' xml <- ocr("https://jeroen.github.io/images/testocr.png", HOCR = TRUE)
#' cat(xml)
#'
#' df <- ocr_data("https://jeroen.github.io/images/testocr.png")
#' print(df)
#'
#' \donttest{
#' # Full roundtrip test: render PDF to image and OCR it back to text
#' curl::curl_download("https://cran.r-project.org/doc/manuals/r-release/R-intro.pdf", "R-intro.pdf")
#' orig <- pdftools::pdf_text("R-intro.pdf")[1]
#'
#' # Render pdf to png image
#' img_file <- pdftools::pdf_convert("R-intro.pdf", format = 'tiff', pages = 1, dpi = 400)
#' unlink("R-intro.pdf")
#'
#' # Extract text from png image
#' text <- ocr(img_file)
#' unlink(img_file)
#' cat(text)
#' }
#'
#' engine <- tesseract(options = list(tessedit_char_whitelist = "0123456789"))
ocr <- function(image, engine = tesseract("eng"), HOCR = FALSE) {
  if(is.character(engine))
    engine <- tesseract(engine)
  stopifnot(inherits(engine, "tesseract"))
  if(inherits(image, "magick-image")){
    vapply(image, function(x){
      tmp <- tempfile(fileext = ".png")
      on.exit(unlink(tmp))
      magick::image_write(x, tmp, format = 'PNG', density = '300x300')
      ocr(tmp, engine = engine, HOCR = HOCR)
    }, character(1))
  } else if(is.character(image)){
    image <- download_files(image)
    vapply(image, ocr_file, character(1), ptr = engine, HOCR = HOCR, USE.NAMES = FALSE)
  } else if(is.raw(image)){
    ocr_raw(image, engine, HOCR = HOCR)
  } else {
    stop("Argument 'image' must be file-path, url or raw vector")
  }
}

#' @rdname ocr
#' @export
ocr_data <- function(image, engine = tesseract("eng")) {
  if(is.character(engine))
    engine <- tesseract(engine)
  stopifnot(inherits(engine, "tesseract"))
  df_list <- if(inherits(image, "magick-image")){
    lapply(image, function(x){
      tmp <- tempfile(fileext = ".png")
      on.exit(unlink(tmp))
      magick::image_write(x, tmp, format = 'PNG', density = '300x300')
      ocr_data(tmp, engine = engine)
    })
  } else if(is.character(image)){
    image <- download_files(image)
    lapply(image, function(im){
      ocr_file_data(im, ptr = engine)
    })
  } else if(is.raw(image)){
    list(ocr_raw_data(image, engine))
  } else {
    stop("Argument 'image' must be file-path, url or raw vector")
  }
  df_as_tibble(do.call(rbind.data.frame, unname(df_list)))
}