How difficult is it to implement image recognition in iOS and Android without internet?

Question

Android

The question is - how difficult is it to implement image recognition from a camera in a smartphone without the Internet? The task is approximately the following - for example, there are 100 vases (the set is known in advance) - they are of different shapes, with different patterns (there are no identical ones, but there are similar ones), how difficult is it to implement an application that will identify a vase from a photo? Is it possible without Internet access at all and how labor-intensive?
After searching the Internet, I found several SDKs that seem to promise object recognition, but because I have not come across them, and I myself am not a mobile developer, I don’t know how they work :-)

Answer 1

Yes, it's possible.
System costs: you need a good CPU and a capacious battery.
Implemented by defining the boundaries of color areas through Gaussian blur.
Some knowledge is required to work with neuro-algorithms for learning.

Answer 2

xmoonlight And how good is the CPU needed? I mean, will only the top-end smartphones pull this, or will most of them work?
And most importantly - is it possible to transfer this data to another smartphone? That is, to make a universal application that will recognize vases from this collection (exclusively them) on any smartphone, or do you need training in each case?

Answer 3

Here's how it was implemented on Android, only in English.