Fecha de publicación: 22/01/2021 Fuente: WIPO Food Portada The invention discloses a method for analyzing food nutritional ingredients in an image based on machine learning. The method comprises the following steps: step 1, acquiring food picture informationuploaded to a server by a user; 2, inputting the obtained picture data into a pre-trained food classification model, and discriminating pictures by the food classification model; 3, sending a result output by the food classification model to a mobile phone APP of the user, and the user confirms whether the food is the kind of food or not; step 4, if the food is the kind of food, enabling the userto confirm and input the weight data of the food; 5, if the type of the food is not the type of the food, enabling the user to manually input the type of the food and inputs the weight data of the food, and learning the type manually input by the user and the picture uploaded by the user again; and step 6, finally presenting a result to the user. Compared with the prior art, the picture recognition network used in the invention is specially trained by using the food data set, and has high accuracy for food picture recognition.