For food safety and consumer satisfaction, it is of paramount importance that food products are hygienic and intact. Indeed, during processing, food can become damaged, become mixed with foreign objects and, at times, be exposed to excessive heat.

In this context, the aim of a recent study, carried out by Turkish researchers (Kılcı and M. Koklu, 2025), was to develop a method for the automatic classification of biscuits based on advanced computer vision models.

During the experiment, a dataset comprising 4,990 images was created, initially dividing the product into two categories: defective and non-defective. Subsequently, the defective biscuits were classified into three subcategories: incomplete, overcooked and with structural defects.

The results demonstrate that the proposed models enable classification accuracy of over 96 per cent in both the two-category and three-category tests. Defective products can then be removed from the production line using laser systems, robotic arms or ejection systems.

In conclusion, the authors highlight that the proposed system can play a key role in automated production processes, increasing production speed whilst reducing labour costs.

Bibliographic references: O. Kılcı & M. Koklu, Food Analytical Methods, 18, 2025, 815-829