DESCRIPTION
The reduction of food waste is a key element within the EU Green Deal, specifically in the Farm-to-fork strategy. The EU Parliament explicitly welcomes the efforts to rework the food system regulations, and (see https://bit.ly/3IMO6U7) calls for a revision of the EU rules on shelf-life date marking that in the future needs to be science based to reduce food waste while keeping EU food safety levels high. The FreshIndex approach and the experiments on mold and fungi of this proposal will be the solution to match this need. The EUH4FreshIndex project will assure that our AI modelling can keep up with customer demands and that the company strategy is well aligned, to leverage this arising opportunity.
Increase accuracy and reliability of FreshIndex AI models
The project focuses on the rapid and non-destructive evaluation for the remaining shelf-life of fresh produce using spectroscopic but also other available data sets.
The key innovation of the experiment is to build AI models for time evolution of measurable quality parameters and subjective quality perception ratings caused by ripening and spoilage mechanisms in fruits and vegetables. Straight forward neural network modeling of storage tests does not provide sufficient accuracy as the detailed properties of many individually different food units are averaged. Through the food scanners these detailed properties can be captured non-destructively with good accuracy and our first shelf-life models show decent results on many cases.
The frequent outliers suggest that by taking into account not only the scanner spectra but also other existing data sets such as the conditions along the food supply chain or details of the harvesting process into account.
Expiry data labelling causes large amounts of food waste, while at the same time over half a million cases of food borne diseases happen across Europe every year. FreshIndex has the vision to provide consumers with true shelf-life information for all perishable food products such as fruits, meat and dairy products. This shelf-life information is continuously updated based on the true storage conditions to which the individual product units are subjected.
The EUH4FreshIndex project focuses in including more and more different, complementary data sets into the AI modelling for food spoilage and product shelf-life. While KER-1 and KER-2 cover the technical maturity for food quality and shelf-life evaluation regarding model accuracy and reliability, KER-3 covers the business and value proposition aspects of the project.
The technical KPI are well aligned with the Key exploitable results of the project, namely KER-1 and KER-2.
