Forum

Reality picture detection

Hey, is there a function/node for reality image recognition such as:

Artivive – The Augmented Reality Art Tool

Regards
Jens

i’m afraid i don’t believe we have anything like this yet in the univvvverse. i guess something could be put together using VL.OpenCV but that might take quite some effort and not work as good as the system shown.

to me it seems that most of such api’s are for mobile only. if you find an api that works for desktop and has a .net api, it can almost certainly be wrapped into a nice set of nodes…

Agreed, I thought about it for a bit and OpenCV does have the feature detection/recognition functionality in there. I am not sure how well it would perform in realtime and also it has not yet been added to the VL.OpenCV wrapper.

What do you think his approach to recognition. is it edge detection and / or object detection?

In this example they work a lot with typo and hard edges.

https://www.behance.net/gallery/79737469/POSTPRINT-AR-3D-in-Graphic-Design

@envoy It probably use a feature tracker like SURF or ORB. There is an example of how this is done with OpenCV here

AS @ravazquez mentions it isn’t clear how well it would perform without GPU acceleration.

@Hadasi thanks for the hints

both functions are very interesting, it would be great to have them in VVVV

as far as i understand the OpenCV source code is already existing

what would be the next step to get out of it an OpenCV.vl for gamma

@envoy, there is already a wrapper for OpenCV in gamma called VL.OpenCV.

You can see more on it here and here.

Unfortunately as I mentioned in my previous comment, the feature detection parts of the API have not been added to the wrapper yet and even if they were, since the library currently works on CPU only performance might be on the slow side.

@envoy, as ravazquez mentions, OpenCVSharp is already available in its raw form. If you want to you can expose all of the nodes that are used to make the VL wrapper and try to make the tracker yourself. You can use the VL wrapper as a bit of a guide as to how it should be done.

Yeah, but Randall referred to the OpenCVSharp wrapper itself, not the VL one wrapping the wrapper :)
So the feature detection hasn’t been added to OpenCVSharp yet.

@readme maybe I was not clear enough in my previous message, I was talking about the VL wrapper itself (VL.OpenCV).

OpenCVSharp does include support for feature detection, at last a few of them. You can check out examples here: https://github.com/shimat/opencvsharp_samples

So @envoy you would basically have to follow @Hadasi’s suggestion and implement the VL functionality you need on top of OpenCVSharp. You could if you want have a peek under the hood of VL.OpenCV to see how many of the common OpenCVSharp tasks have been tackled already.

Another option is to “extend” VL.OpenCV in your own custom VL document which references both VL.OpenCV and OpenCVSharp, giving you access to what is already there in VL.OpenCV while also letting you access lower level OpenCVSharp member as you see fit.

In VVVV.Packs.Image there is a SURF node DetectFeature (CV.Image).
When I open the help patch the “ImageToProjection” node is missing, someone knows where I can download it