Does the computer vision modules really work at image detection?

Hi folks,

I’ve been a long time user of UI Vision from the Kantu days, but just this past year, I’ve been trying to use the XModules / computer vision to interact with some Google sheets dialogs that are not traversable via the DOM. And, after many failed attempts at recognizing basic images, I just gave up. Every time I try to look for the same image in the same location, the computer vision modules will bizarrely find other areas of the web page and click on random areas. It got to the point where I just had to hard code the X , Y coordinates into XClick, because that was the only reliable way of doing this. However, that ran into more issues as UI Vision wasn’t able to recognize when the task was completed, and would then just proceed thereby shooting itself in its own foot. I had to hardcode a timeout, hoping the task would complete before it would start the next iteration of the loop. I then just gave up entirely when I realized XClick allows on 25 clicks for the free version.

Anyways, since then, I found Pulover’s Macro Creator, and IMO, it’s better than UI Vision. It’s able to recognize images properly, and I can build good logic around that to make my script run smoothly with little interaction from me. Not only that, but it’s entirely free. But, it has me wondering, why is UI Vision computer vision components so terrible. I can do a screen cast record to show people what I mean. Does it actually work for other users?

Hi, it works well for me! And since version 8 uivision can even do OCR locally.

Anyway, I would love to see a video or a screenshot of where the UI Vision image recognition does not work for you. I am not saying it is perfect, but it in my of use cases.

wasn’t able to recognize when the task was completed

Why not? And “Pulover’s Macro Creator” works better?

PS: The Pulover’s Macro Creator website is full of spammy ads, I stopped browsing it after a few minutes. That does not inspire confidence in the future development of this tool. That is the problem with free tools, at some point they just get abandoned. But if you like that tool, would using AutoHotKey (AHK) be not even better?

Hi, sorry for the late reply. I’ll try to post a video sometime next week about the issue I’m facing. But, as it is, I’m quite busy.

I like UI Vision because I am not forced to use image recognition but can recognize web elements instead. When I need to add a report to download to my macro which does this, I just add it to a csv file and no need to screen capture anything.

I was looking at the first tutorial for Pulover’s Macro Creator and it didn’t indicate the ability to do that.

I personally didn’t notice any ads on the site but that might be due to UBlock Origin add on I use.