Federal learning-based food material identification method and food material identification model training method

• Main Authors: XIE FEIXUE, LU SHUXUAN, LI ZHENGYI, ZHAO QIDONG, GAO YUHAN, QU LEI
• Format: Patent
• Language: Chinese; English
• Published: 23.05.2023
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

The embodiment of the invention provides a federated learning-based food material identification method and a food material identification model training method, and the method comprises the steps: carrying out the training of a food material identification model which carries out the food material identification locally through a federated learning architecture, and uploading the gradient data obtained through training to a server; and taking the server as a federal learning participant to carry out aggregation training on the gradient data of each local identification device to obtain model parameters, and issuing the model parameters to each local identification device to update the food material identification model. In this way, the food material images shot by the local identification devices do not need to be uploaded to the server, so that the privacy of the user is protected. And a federal learning architecture is adopted, so that the food material identification model obtained by training of each lo