“Calibrate - OCRTEXTX” (UI.Vision RPA → Settings → OCR) doesn’t seem to have any documentation in the User Manual and it doesn’t return any results when searched in the forum or on google.
My main question is: what is the number input for the “Calibrate - OCRTEXTX” intended to do?
I also did some experimenting with the “Auto-Calibrate” workflow to see if I could figure out the relationship on my own, but I ended up uncovering some (to my understanding) strange behavior, and I also have some general questions about the Auto-Calibrate process. I don’t want to muddy up this topic post with long documentation, so I made it a shared google document: Calibrate - OCRTEXTX_Behavior Documentation_UI Vision - Forum (Please let me know if I should append this Topic with that content, or if any of those documented behaviors should be submitted as their own “Bug Reports” tagged Topic, and I’d be happy to do so!)
I’ve been having issues with inconsistent OCR performance (offline and online), so I wanted to start by learning more about the calibration to make sure it isn’t my setup that’s causing the variable performance.
My setup info:
Windows 11, version 22H2
UI.Vision RPA 8.3.2
LOCAL OCR Processing (v1.0.12)
(and my XModule root directory is in the default location on the Desktop (ref))
First of all, let me clarify what gets calibrated. When you have e. g.the command XClickText| abc#R20,-10 the RPA software puts the click position 20 times the size of “x” to the right, and 10 times the size of “x” down (negative = down, position = up). And for this, we need to know the font size of “x” in pixels. Usually a calibration value of 6, 7 or 8 is correct.
The calibration value is used for these commands:
XClickText
XClickTextRelative
XMoveText (available soon)
XMoveTextRelative (available soon)
OCRExtractTextRelative (available soon)
All this applies only to desktop automation OCR. For web automation mode, the calibration is not needed and not used.
we need to information on the font size, so
Is “Windows Scaling %” supposed to match my Windows OS instance?
Thank you so much! Your response is extremely helpful, and I’m glad my feedback format was well received.
I’ve been exclusively working in web automation mode, so I think my OCR issues are unrelated to the OCRTEXTX calibration, but this will be really useful when I start automating desktop processes.
Now that I know this probably isn’t related to my OCR struggles, I’m going to do a bit more experimenting with different OCRExtractRelative implementations before I consider creating/contributing to another Topic for some help. Is it possible for me to use the “abc#R20,-10” format the target value for an OCRExtractRelative command? Or is that method of OCR extraction going to become available upon release of the OCRExtractTextRelative command?
My last follow-up question is regarding your use of “abc#R20,-10” as the target Value for an “XClickText” command: what is the difference between using the command “XClickText” with the coordinate string, compared to using “XClickTextRelative” with that same value?
And thanks again, I really appreciate the help!
P.S. For any future forum frequenters: I found that there was already documentation for the OCRTEXTX calibration in the XClickTextRelative section of the XClick documentation page. The explanation provided in this thread by the admin is great and helped me find it in the documentation–the preexisting explanation just provides an additional example that could be helpful!
All this applies only to desktop automation OCR. For web automation mode, the calibration is not needed and not used.
- Obviously this will be fixed with the next update.