CV Click - Is it normally this slow?

Help Activities
,
1

Hi everyone,

The computer vision activities in my process are taking a long time to complete. I’m using CV ElementExists and CV Click. In the worst case, these two activities take over 4 minutes to complete, even though the visual elements are on the screen within about 20 seconds.

Is this normal and, if not, how do I begin to investigate what’s causing it?

Thanks for any help

Edit: The CV activities are finding & clicking an icon within a VirtualBox VM - the typical UiElement activities don’t work for this, so CV is my last option.

1 Like
2

CV activity being image based activity it is very obvious that it will take time
But we can reduce the time taken with few changes in properties like Timeout property
The default timing is 30 seconds which we can reduce to 10 or 20 to make the bot wait only till that timing and if not found it will give exception

But I would suggest to keep it in default and it’s not a problem in taking time unless it works fine as expected

If it really matters in tuning performance then we can try with HOT KEYS which is useful if it’s a Citrix or images based environment where we can access the elements by navigating with tab key or right left keys

Cheers @rpa_jay

3

Which CV server are you using? The default one (our cloud server) or something else?

4

Hi @Cosin,

I’m using the default UiPath server. Are there alternatives that might be faster?

Thanks

5

No but if you were using an on-prem one that could be the source of the issue.

I’ve sent the link to this thread to the devs, they might require some more info from you to investigate

6

Hi!

Any chance that you are using LocalServer?
Do you have any other automations that is running at that moment with that ApiKey?
What is the resolution of the screen / app?

Thanks,
Stefan

8

Hi everyone,

The CV activity is using cv.uipath.com as the server.

The same ApiKey is used in prod and test, but I couldn’t find any instances where the two would have run at the same time.

The robot is running with 1920x1080 screen resolution, but the actual VM element is 1024x768. Is this too many pixels?

Thanks everyone

9

Hi,

That sounds that it should work ok, much faster than 4 minutes.

Is it possible to install Fiddler and capture the calls (with https decryption enabled in Fiddler) and share the logs?
Also would it be possible to share the Workflow?

Thank you,
Stefan