JoelSPendery
commented
5 years ago I found a work-around to this issue. Basically, I'm running a batch file within the script that kicks off a new instance of the script after I complete the OCR of a document. Immediately after kicking off the batch file from within the script, I run a quit statement to the previous instance of the script which releases the memory. It's not elegant, nor does it fix the root cause, but it works.
shell('C:/location/of/bat/file', wait=FALSE, translate = TRUE) q()
Batch file:
"C:\Program Files\R\R-3.5.1\bin\Rscript.exe" "C:\location\of\r\script.R"
You will need to have some way that you search (and continue) through the documents that have already been run through OCR.
Oneiricer
I am reading in a number of pdfs, performing pre-processing and then getting the text from the ocr_data method (from a temp PNG file). During my loops to go through each pdf and to get all the pdfs in a directory, I am removing unused variables and performing garbage collection (gc()) to reduce the memory load of my program. However, as I continue to read in more and more pdfs, the memory load on my system keeps increasing, eventually running out of memory and throwing an error. I can load the library and comment out the ocr_data() method call in my script and the memory load doesn't appear to increase (minimum load is fairly consistent over a time period where I have been observing an increase). If I replace ocr_data() with ocr(), I don't see an increase in memory load on my system. This would suggest that whatever is causing this issue originates with the ocr_data() method (probably something isn't being deleted on the C++ side that might be related to saving the data frame values that are returned, but it's just a guess).