AI Computer Vision is Now Available for Preview

[sachin_sharma]

there are many awesome feature i will try it:clap:

[KevinS (Kevin Stoneham)]

SUCCESS!!

I’ve tried the beta release and found it to be a big improvement in a previously difficult (but common) use case. Great job team. Here are the details:

I recently produced a demo video called, “Can a Robot Do My Taxes?” DigiServ RPA Demo - Can a Robot Do My Taxes? - YouTube

As an initial step, the automation reads PDF tax documents such as a W2, which have a dense plane of text in boxes with very small header type. I ran into the well known issue where Get Text and similar activities have difficulty recognizing a single data field within such a PDF. I had to try many different approaches before finding one that was suitable. Even then it was bulky with many additional activities required.

After this week’s announcement I loaded Computer Vision and was able to easily grab the single bits of text that I wanted from the W2 - no special config steps, no string manipulation required. Better yet, I ran other W2 examples through and it remained accurate without adjustment.

Again, great job! Can’t wait for this to go into production.

Kevin

[Ryan_Rush]

Really excited to hear this! Thanks for letting me know.

[Cosin (Cosmin Voicu)]

thanks kevin! it’s awesome to hear new/unexpected ways this is useful

Not sure if you’ve tried a somewhat hidden feature: the box-selection for GetText. It can scrape a designated area relative to an anchor. Maybe you’ll find it useful

[html12345 (jeni)]

Thnks for the post

[KevinS (Kevin Stoneham)]

Thank you @Cosin for the tip. I don’t think I’m familiar with the box-selection feature. Is that part of the original Get Text activity or the new CV Get Text in Computer Vision? If there’s any documentation on this feature I’d be interested, thanks!

[Cosin (Cosmin Voicu)]

Part of the new CV set of activities.

To try it out, in the CV.getText, click indicate and do a box selection in the wizard, instead of clicking a word. After that you will need to pick an anchor for it. At runtime, it will scrape the whole area of the selection. Very similar to scrape-relative

[Kadiyala_Venu_Madhav (Kadiyala Venu Madhavi)]

thanks

[Anas.Mohammed]

excellent

[ImPratham45 (Prathamesh Patil)]

I am facing error while running CV activity in beta version

Please help me out to resolve this issue I am very excited to work with CV

[loginerror (Maciej Kuźmicz)]

Hi @impratham45

Please restart your machine

[ANSHUL (MANTRI)]

thanks

[chetan8626 (Chetan R)]

Thanks a lot

[dmitri47 (Dmitri Mahayana)]

Thats awesome.
Good job UiPath Team

[ASinha]

Great feature and excited to use.
But I’m getting this error pop up message, what should I do?

[Cosin (Cosmin Voicu)]

Please see first post with installation instructions

[harlonleehugo (harlon)]

Nice Stuff, issue on Tables blank selection, but still amazing

[Makoto_Tamura]

Can AI computer vision correctly recognize Japanese UI elements?
In my environment, AI Computer Vision could not correctly recognize the button with Japanese written on the label.
If AI computer vision does not currently support Japanese, are you planning to support Japanese?