Automated Assessment of Pizza Quality Using Computer Vision and Deep Learning Techniques

Quality control in the pizza-making process is essential to meet consumer expectations. In the age of online delivery, verifying the delivered pizza is necessary. The pizza usually consists of a base, sauce, cheese, vegetables, and meat. Low-quality pizza delivery may lead to consumer dissatisfaction and hinder pizza production business growth. Hence, this paper proposes a method to evaluate sauce spread, identify toppings, and assess their distribution on cooked and uncooked pizzas to address this issue. The proposed methodology uses a feature pyramid network for background and foreground separation in the pizza image. Then, the sauce spread thickness is classified into 3 categories using the U-Net model. Finally, a multi-label CNN and color segmentation were applied to the image to evaluate the spread of toppings on the pizza. Experiments were carried out on a standard image dataset, and it was observed that the proposed model achieved 95% accuracy in pizza quality evaluation.