How to organize the search for objects in the picture?

Question

Neural networks

Let's say I took a picture of my face. At the same time, I really needed to find out the location of the nose on it, just so that it would be framed. Tell me what (Well, or reading on this topic, I will also be grateful) in general there are ways to implement this.

Answer 1

There is a well-known annual ImageNet competition, and one of the tasks there is object recognition / localization in an image (participants are given a huge database of labeled images). In 2015, the Microsoft team's network won in the detection/localization section ( results ). Their article on arxiv is here . And here is another article by the same people.
In general, when it comes to doing something with an image, you should always use convolutional deep neural networks. Because the picture is a fertile ground for extracting the hierarchy of features, and convolutions help to reduce the space of features for higher logic (whatever they are).

Answer 2

Start with OpenCV for example...