Fecha de publicación: -- Fuente: WIPO "tomato" The present disclosure provides a computer-implemented method for determining a nutritional value of at least one dish. The method comprises providing a plurality of images of the dish taken from different perspectives; using a first machine learning algorithm to detect the dish type from the images; and processing each image with a second machine learning algorithm to determine the dish's volume. The second algorithm includes a shared Vision Transformer (ViT) encoder for each image to generate feature representations, and a ViT decoder for each image to produce a point map and a confidence map. The decoded point maps are converted into a three-dimensional model for each dish, from which the volume is estimated. Nutritional information is retrieved from a database using the detected dish type as a key, and the nutritional value is calculated based on the estimated volume and the retrieved information. The disclosure also provides additional independent claims, including a method for calculating the nutritional value of an eaten portion of a dish by comparing before and after consumption images; an electronic device configured to perform the methods; a portable electronic device with a camera unit for capturing images and interacting with the electronic device; a system comprising such devices; and a computer program product comprising instructions to implement the methods.