How to implement photo recognition on js and neural networks?

Question

JavaScript

There is a task to implement in the browser recognition of the device in the photo with the output of information after successful recognition ... The bottom line is that there are enough libraries where you can write, for example, a number and AI will determine it. If trained enough.
Need something similar. Only recognize, say, a printer that the network already knows about. Or a scanner...
It will be a web application. I have been studying networks for a couple of weeks...
1. How to teach the network to recognize them? I know that there are special solutions. Found only for face recognition. and it does not fit (
2. How to train the network to determine the model of the device from the pictures?
3. It's just fine if there is one with a clear manual (and then I woke up, realizing the harsh imperfect world of the jun =) ) ) ?

Answer 1

The recognition of digits is done on the basis of training the network with a rather large set of MNIST pictures, where each drawn digit corresponds to its correct value.
For JavaScript, take the ConvnetJS library for example .
Do you want a ready-made network trained to recognize computer equipment from pictures? You hardly find it.
Write and teach yourself by parsing some computer equipment store - it is easy to compare the picture and the device class corresponding to it. But the set of only sold models is probably too small. You can also add Yandex/Google search results based on pictures.