A virtual taster has appeared in Russia that doesn't need to taste sweets. Specialists from MIREA Russian Technological University have developed a recommendation system that uses a neural network to evaluate the future taste of a dessert based on its production data. If something goes wrong, the system can quickly detect it and point out a possible defect.
Currently, the quality and taste of finished products are checked by a tasting commission. Specialists taste the products and then give their conclusion. Such a check takes a lot of time and does not allow for quick adjustments to the production process.
The virtual taster, in turn, receives data from various stages of production — from sensors and the enterprise's laboratory. The system takes into account the exact proportions of ingredients, cooking temperature, consistency, color, smell, and appearance of the product.
Based on this data, the neural network evaluates the taste of the finished dessert on a 10-point scale. If the rating is low, it means that the product does not comply with the recipe or the technological process was violated. For example, if too much sugar was added to the product at one of the stages, the system will record such a deviation.
The development is based on a neural network model that was trained on real data from confectionery productions.
Taste is a subjective category, but it can be predicted if based on precise numbers. We have created a model that links technological parameters with final taste characteristics. It's like a recipe where the ingredients are not grams, but entire technological regimes.
According to him, the developed approach is similar to a recipe, only instead of just grams of ingredients, entire technological regimes are taken into account.
As a result, the technologist can see the predicted taste of the product even before the first batch goes to packaging. According to Blagoveshchensky, this saves time and raw materials.
Read more on the topic:
- Packaging itself will show that food has spoiled: a Moscow schoolboy created an indicator film
- Chef Robot: Nearly 40% of Russians Willing to Try Food Prepared by Machines
- The first AI cashier for cafes and restaurants was developed in Russia