CV license question

Hello all, I am struggling with a weird issue in our production environment where for a particular process we have added 3 bots to keep up with the volume. This process uses Computer vision as it deals with applications behind citrix. We have a Computer vision license hosted on a remote server, that the bots uses, Of late we are seeing very low productivity (successful transactions) up on addition of the multiple bots in the environment.
My question is, do we need multiple Computer vision licenses with respect to the increase of number of bots working in the environment? Will that have any impact with our number of successes? The bots are using the URL where the CV licenses are hosted and are not explicitly using the API key in the code or orchestrator assets.
Can anyone through some light here? Thanks in advance.

Hello @Anna_S!

It seems that you have trouble getting an answer to your question in the first 24 hours.
Let us give you a few hints and helpful links.

First, make sure you browsed through our Forum FAQ Beginner’s Guide. It will teach you what should be included in your topic.

You can check out some of our resources directly, see below:

  1. Always search first. It is the best way to quickly find your answer. Check out the image icon for that.
    Clicking the options button will let you set more specific topic search filters, i.e. only the ones with a solution.

  2. Topic that contains most common solutions with example project files can be found here.

  3. Read our official documentation where you can find a lot of information and instructions about each of our products:

  4. Watch the videos on our official YouTube channel for more visual tutorials.

  5. Meet us and our users on our Community Slack and ask your question there.

Hopefully this will let you easily find the solution/information you need. Once you have it, we would be happy if you could share your findings here and mark it as a solution. This will help other users find it in the future.

Thank you for helping us build our UiPath Community!

Cheers from your friendly
Forum_Staff

Hi @Anna_S

A couple of questions:

  1. I understand you are using an On-Prem CV Server. Is it Windows or Linux?
  2. Is the On-Prem CV Server using GPUs or CPUs? (a hint here would be if a simple click on a fullHD screen takes 1-2 seconds or 9-10 seconds)
  3. Any issues with HDD space / RAM / GPU RAM on the CV Server?

Best,
Stefan

Hello Stephan, Thank you for your reply. Below are the answers to your questions:

  1. I understand you are using an On-Prem CV Server. Is it Windows or Linux? - Windows 10/Azure VM

  2. Is the On-Prem CV Server using GPUs or CPUs? (a hint here would be if a simple click on a fullHD screen takes 1-2 seconds or 9-10 seconds) - CPUs
    image

  3. Any issues with HDD space / RAM / GPU RAM on the CV Server? – No issues on the VM that we are aware of.

All we want to know is with the increase of number of working bots simultaneously on a process, do we have to increase the CV licenses as well? Or what are its impact on the productivity(successful transactions) that involves CV activities and image matchings…

Hi Anna,

The CPU server is quite slow (probably 5-20) seconds compared to 300- 500 ms with GPU) and what’s probably happening is that when you only had 1 robot it was sequential and never waiting.

Now (probably)
0 seconds Robot 1 sends an image (it takes 20 seconds)
5 seconds Robot 2 sends an an image (it takes 20 seconds, but starting after the previous is finished)
6 seconds Robot 3 sends an image but it never is finished before the timeout of 30 seconds.

What I would suggest is testing LocalServer option for the three robots (together with ComputerVision.LocalServer package) and see if that solves your issues. The LocalServer uses CPU but on each and every robot and it might be exactly what you need.

The latest LocalServer package + UIAutomation 21.10 also have UiPathScreen OCR capabilities.

1 Like

Hi,

Thank you for your suggestion.

Wondering what is ComputerVision.LocalServer package? I am new to this.

Questions:

Are there any code changes that needs to be made if I download this new package (ComputerVision.LocalServer package)?

What do you mean by testing LocalServer option for the 3 bots? Does it mean you want us to install the CV license on each of these bots runner machines or have that installed on a remote server which is not on cloud/azure.

The automation is developed using the studio version (2020.4.1)to run unattended using the Orchestrator version(on prem) : 2020.10.3 . Wondering if this new package (computervision.localserver package is compatible with our studio and orchestrator versions?

It will be great if you can share any documentation around this.

Hi,

First the link:

ComputerVision.LocalServer is a new(ish) way of using ComputerVision. Instead of calling an outside server (Cloud or OnPrem) we use the CPU on the robot machine to run the model and detect the controls + text.

To use you need to set each and every CV Screen Scope (or directly in project settings) the Use Local Server to true. You will also need to install the ComputerVision.LocalServer package in your project. If you also want to use UiPath Screen OCR with Local Server you need the UIAutomation 21.10 package and to set the UiPath Screen OCR Use Local Server property to true.

Afterwards you can publish the updated project and deploy it to your robots. What might be needed on the runner machine is the Visual C++ Redistributables for 2015-2019 , but nothing else (as per the troubleshooting in the link above)

Your Studio / Orchestrator versions should be ok with updating UIAutomation / installing ComputerVision.LocalServer.

Please tell me if this helps,
Stefan

Hi @sbotan ,

I’m currently evaluating the Computer Vision capability of the platform on a PoC.
I understand that the cloud API version is an add-on for Enterprise license at an additional cost.

For the Local Server Pack, once the package UiPath.ComputerVision.LocalServer is downloaded for the project, it can be used without the API Key/URL.

Just wondering, does that mean there is no extra cost/consumption limit associated with this approach? Or is it somehow still limited by the xxx Megapixels/Min consumption as per the cloud endpoint version?

image

Hi @warren_lee .

The LocalServer option does not have extra cost / consumption, but it will be slower than the cloud version, as it’s being run on CPU, instead of GPU.
Also it might have less accurate detection and at the moment it does not support Table detection.

Hi @sbotan ,

Thanks for the swift response!
Gotcha, that clarifies it. This can be a good option for clients with potential license constraint but is still wanting to leverage the capability of computer vision of the platform to certain degree.

Totally understand the less accuracy given it is probably a smaller model running on pure CPU.

Just wondering, if it’s on the roadmap for the local CV model to be improved so that it’s able to deliver all CV capabilities that the cloud version can do (apart from the accuracy part, e.g: CV Get text + table detection)?

Hi @warren_lee, yes, we’re aiming to fully deliver this in 22.10 - will release a new localserver model with table detection in 22.6 and working on adding auto-anchors as well for 22.10, thus making it fully on-par feature wise with our cloud model. Still slower and slightly less accurate (it is a compressed model, as you mentioned), but it’ll be on-par.

Also, remember you can always use the Community cloud license if 30 MP/min will do, also for free.

1 Like