0

I need to create a neural network that will be trained using a set of pictures, for example 1000 pictures. Then I want this network to be able to take a as an input a video from camera and detect if it sees one of these pictures - but not on entire screen, but for example as a printed picture on the wall. And from it I would like to get some most probably seen pictures. I don't need to know their location in the input picture, just an information about which are they. So this network would not be a classifier - Im not interested in what this picture is of - just a name of this picture, or index in the set, or whatever. Is there some kind of neural network capable of doing something like this? It can be TensorFlow, CoreML or MLKit or whatever else.

Andy Jazz
  • 49,178
  • 17
  • 136
  • 220
Damian Dudycz
  • 2,622
  • 19
  • 38

1 Answers1

0

try start from CreateML utility.

Xcode -> Open Developer Tool -> CreateML.

Then Select "Object Detection" project.

It's used YOLO2, it's have good performance but might be not so accurate.

I recommended try it first, if this solution can't cover your's requirements then I will try help to you with YOLO5.

Dmytro Hrebeniuk
  • 124
  • 1
  • 6
  • Also, you can try alternative way, using ARKit, [link](https://developer.apple.com/documentation/arkit/content_anchors/detecting_images_in_an_ar_experience) ARKit picture detection sample – Dmytro Hrebeniuk May 05 '21 at 18:18
  • Thanks, but object detection detects certain types of objects, not a specific picture. The difference is that I don't need to know that I see a tree, train, car, etc, I just need to know that I see "Image 003.jpg" - exactly the same picture, and not some specific object in any visual form. – Damian Dudycz May 05 '21 at 20:33
  • As for ARKit image detection, this is what im going to use in next step, but it only works fine for certain number of pictures. That is why I want to help it with some initial information, to limit the number of pictures it needs to search for. – Damian Dudycz May 05 '21 at 20:34