DoDo Pizza started the neural network controlling quality of the made pizza

On June 19, 2018 it became known that DoDo Pizza started neural network which will allow to control quality of the made pizza in any institution of network. The application on the basis of machine learning was developed on a blockchain platform for creation of artificial intelligence of Dbrain.

The application works in the form of Telegram bot. The bot was developed for 'army' of mystery shopping - it is the internal community of DoDo Pizza numbering already more than 50,000 people. DoDo Pizza creates the rating of quality of pizzerias according to photoreports of clients, and until recently only living people who weekly estimated nearly 700 pizzas on compliance to standards acted as experts. Now checking will be able to send the photo of pizza at any time to a bot which will at once give an assessment to a product from 0 to 10 by key criterion — quality of the test.

The neuronet makes pizza

For the application which will recognize quality of the test on the Dbrain platform more than 50,000 photos of pizza made from different foreshortenings were marked. The command developed model of neuronets which passed several grade levels for the solution of this task. So, at the first stage it segments pizza and finds defects. For example, if at edges of pizza small swellings which are almost not noticeable to a human eye — it at once speaks about bad quality of the test. At the second stage the model studies at data which were provided by the mystery shopping and experts who are engaged in pizza quality evaluation. So the neuronet understands a pizza quality assessment system people and can help them with the solution of this task.

Training and testing of Telegram bot took 1.5 months. In the course of work the bot addresses the platform and transfers data for the analysis. DoDo Pizza calculates that the bot will help to reduce quantity of errors, and in the future completely to replace the person and many times to increase amount of the checked pizzas. For example, using the video camera on the oven the program will manage to check not 2 pizzas in a week, and 300 in day — i.e. all pizzas which are going out of the furnace.

Together with Dbrain we managed to create the tool which helps to determine quality of the test by reports of mystery shopping already now. We want to improve further technology to have an opportunity to check each pizza which is going out of the furnace. It will help to hold at height quality of pizza at further scaling worldwide, reducing at the same time costs for quality control. It only the beginning. Fedor Ovchinnikov, founder of DoDo Pizza