VM's broke our Computer Vision?

Hi all!
So we just stood up 3 new vm’s at work. They’re just regular Windows 10 machines. We made sure the settings on all the applications we use match what’s in our dev environment, and so far the bots can log in and work just fine.

Until we use computer vision.

Every single project that uses computer vision has failed. We haven’t had issues with CV on our VM’s in the past, but we were also using a Windows 10 Server with Terminal Services on it, not just a standard Windows 10 machine.

Has anyone had this issue before? We’ve found some success in going in and manually reindicating each CV window, but we have some processes that are pretty CV heavy and would like to avoid having to do that for everything.


Hi, can you give me the exact exception that you are getting? Also the type of cv context (cloud, on-prem or local server).

If everything works after re-indicating, the reason should be one of the following:

  • selector is now different, so CV can no longer attach to the window as it was described in the initial selector;
  • image is different in some way (resolution, scaling, element placement inside of the app window) such that CV can attach to the window, but it can no longer identify the targets generated for the activities inside of the CV Screen Scope.

Let me know and we’ll work from there to determine how we can fix this in the most efficient way possible.


Hey Drago, sorry work’s been crazy busy. The exact exception was a Selector Not Found exception on the first CV Get Text we use. We’re using Cloud (I believe) with the default UiPath CV option. We’re thinking it’s either a resolution problem, or that possibly the windows in the VM aren’t opening all the way (although the ones we have checked did open).

We did solve the problem on one project. We have a process large enough that we had to clear the screenshots off of, as it was too large to publish to orchestrator. It still worked on our Terminal Service VM, but broke on our new machines. Once we reset those screenshots it worked. That being said, we haven’t deleted the screenshots from our other projects, so the issues on those are different.