How to recognize an object in real time using window capture?

Question

JavaScript

I'm interested in the simplest way to recognize my image on an HTML page in real time. I take any system window with the desired image, display it on the page as a canvas so that the neural network can determine it and draw something on top of the original image. I found face-api.js , which is perfect for me (I implemented it), except for one thing: I do not need to recognize faces, but my own object, which I will set to recognize . I don't know how to do it. Is it even possible now? If possible, how can it be done using JavaScript ?

Optional: (preferably this should only work in JavaScript)

At the moment, I've tried using other libraries, likeml5.js , but they don't fit because they only use the camera and something else. Or I just don't understand how to use them together with the Screen Capture API

Any help would be appreciated!

Answer 1

I have a stupid idea =) Convert the desired image and sample to svg and, accordingly, find part of the sample code in the source.

Answer 2

you need to recognize not faces, but your own object

in general, the problem statement is so-so. Opencv + yolo to the rescue. If the object is specific, you will most likely have to teach your neuron.

Answer 3

I figured it out myself.