Project
Demo application created for recognizing objects offline
Goal
To check the technological feasibility of creating an application that allows recognizing products on store shelves offline.
Obstacle
To create the application, we used a ready-made model of a neural network, which covered not only the goods of the store, but also other objects. Accordingly, many objects were redundant. In such a situation the offline operation of the app was only possible with a limited range of goods.
Solution
We developed an iOS application that allows you to recognize objects online and offline with a certain percentage of accurate recognition probability.
The offline neural network model supported only a thousand objects. As a demo model, the application has completed its task.
For the application to operate offline on a full-scale basis, further training of its own neural network model is required.
