It’s my first experience with ML, and at that moment I created a convolution neural network on pytorch using the pretraining model VGG16 and added my own extra layer to define belonging to one classes. In my case, bee or not.
The model works well. Now I want to train the saved model, so that she determines the amount of identical objects (bees) on the photo.
But I could not find documentation on how to do such things. For this are used CNN or other methods? If it is not difficult to describe the approximate steps, or give links to examples I will be grateful.
You should use another architecture, an object detector. It’s nothing but a CNN which generates a bounding box around learned objects.
This way you can count the amount of bb, thus, amount of items of each class.
Maybe you could give a link to examples with similar networks, so that I would have something to start?
You can have a look at Yolo V3, RetinaNet or SSD.
They have different performances but if you are begginer Yolo version 1 may be a good starting point 
Of course there are many others.
You may also google it and you will find for sure fancy tutorials.
This is a very good site to learn computer vision: https://www.pyimagesearch.com/2018/11/12/yolo-object-detection-with-opencv/ and they did a tutorial about yolo with code and explanations.
Thank you very much for the help and advice.