OCR accuracy on unattended remote

ocr
robot

#1

Hello,

I’m running a robot on a remote machine Windows Server 2012 R2. OCR Reads fine when I run the robot with remote desktop opened, but when it takes over the session and runs on it’s own (using login to console: true)
OCR accuracy is totally garbage.

The robot is opening adobe pdf reader and doing some relative scraping. What could be the reason for this behavior? Robot resolution and depth is configured to be same as when I connect through remote desktop client.

Please help because OCR accuracy is the key here.


Running unattended robot on VM
#2

Hi tjod,

Please try to change the Profile and Scale options of the OCR engine.


#3

Hello everyone,

What I want to achieve: Start the process and have it running when I close the Remote Desktop Connection. The process involves clicking, finding images and OCR activities. No Orchestrator involved.

UiPath version: 2017.1.6435 CE
Machine: Windows Server 2012 R2 Standard 64-bit

What I tried so far:

  • Login to console: false, Launch Workflow Interactive starting after I close the remote desktop connection:
    (0x80070520): A specified logon session does not exist. It may already have been terminated.

  • Login to console: false, Launch Workflow Interactive starting before I close the remote connection:
    Fails on first find image activity (timeout).

  • Login to console: true, using Launch Workflow Interactive:
    Kicks me out of session, still fails at first find image activity (timeout).

The workflow runs fine when Remote Desktop Connection window is open and not minimized. Robot resolution and depth configured both in UiPath settings file and directly in the Launch Workflow Interactive activity. It seems to make no difference whether I start the project from studio, tray or cmd. I put in the generic credentials in Windows Credentials Manager. The account on the machine I try running this on is a local administrator account.

Is it a Community Edition restriction or am I missing something? Would be good to get some clear information from UiPath team, because most of information I could find here seems to be deprecated or simply not working for me.

Please help.

EDIT: I somehow got it running using login to console: true, but OCR accuracy is garbage now, any hints about that?


#4

that’s not it, I think it’s a resolution issue, even though robot is configured to run in 1920x1080 it seems like it runs in a different resolution. is there any way to check the resolution on runtime?


#5

It’s a Resolution issue.

Try kicking it off with Launch Workflow Interactive, with 1920x1080 32depth.

EDIT: Also, you want Login to Console on false or unchecked in that activity. The reason is because when it uses the console, you can only run one robot at a time for those that use the console.


#6

Sorry, I see you also mentioned that it fails on Find Image. That could be another issue. You could try Taking a Screenshot at the moment before it keeps failing to see what the screen looks like. I also normally need to lower Image accuracy down pretty low but not too low where it picks up wrong images.

Regards.


#7

hey Clayton, thanks for your response,
i just checked the resolution on runtime with a vb script and it’s 1024x768 even though Launch Workflow Interactive is configured with 1920x1080 32depth. Any ideas how to overcome that?


#8

Do you have maybe a screenshot or something so we can see how you are using it? Normally, I will call the entire Main workflow with the Launch Interactive, and use the regular Invokes in the main.

If your Launch Interactive is set up with correct resolution then I see it odd that it’s not running as such, so I’m wondering if you are using it incorrectly. :man_shrugging:


#9

Yes, I have the main workflow wrapped inside Launch Workflow Interactive. I don’t think there is something wrong with that, i found following information:

If LoginToConsole is true, the resolution is limited by the graphical card of the robot machine (which might be as low as 800x600 in some cloud environments), otherwise the resolution for a simulated RDP connection is restricted by RAM (which would generally be more permissive).

Could it be that VM settings are keeping the resolution cap at 1024x768?

Turning off login to console won’t work because I can’t even get the process running then.


#10

One thing you could try is RDP in with the same resolution as it seems to be running in, then reselect the image so it runs at the other resolution.
image

Regards.


#11

Hey, I don’t quite understand, I’m able to go in at 1920x1080 with Remote Desktop Connection, the resolution changes when I run the workflow and robot takes over and kicks me out of session.


#12

Sorry, I meant to Remote in with the 1024x768 resolution by changing the settings before you connect. Then, choose the Image for the Find Image activity within that resolution so it can recognize it at the 1024x768 one.

I’m still not sure why it’s not working for you but was providing a possible workaround by designing your workflow with the other resolution.

Regards.


#13

Thanks for your response again, yes that would be some solution but I think it will be better if VM team configures the machine to allow higher res in console session.

Have a good day