Fatter R-CNN food classification and GL value identification method based on attention mechanism SENet

Abstract

The invention discloses a Faster R-CNN food classification and GL value identification method based on an attention mechanism SENet. The method comprises the following steps: acquiring a picture whichis uploaded to a server by a user and contains a reference object, and performing food detection and identification on a food picture through a Faster R-CNN algorithm based on an attention mechanismSENet; enabling the attention mechanism to enable the feature extraction network to use global information without being limited by local small-view-field information, and meanwhile the problems of gradient disappearance and degradation caused by increase of the network depth are effectively avoided; inputting the type and position of the food and the position of the reference object into a trained linear regression model, and predicting the volume of the food; calculating a GL (glycemic load) value of food according to the category of the food and a GI (glycemic index), judging the edible category of the food, and generating an edible suggestion of the food; and finally feeding back a result to the user.

Description

technical field [0001] This application relates to the field of food safety, the field of image recognition, and in particular to the calculation method of GL value. Specifically, it is a method for predicting changes in human blood sugar load value caused by dietary intake during meals. It mainly provides blood sugar load reference for diabetic patients, establishes and improves diet structure, and forms a virtuous circle of convenient diet. Background technique [0002] Introduction to existing technology [0003] Diabetics who want to measure blood sugar need to wash their hands with warm water and neutral soap first, and rub the finger that needs blood collection repeatedly until the blood supply is rich; use a blood collection pen to stick to the finger pulp, press the spring switch, and acupuncture the finger pulp; turn on the blood sugar Switch on the meter, take a test strip and insert it into the machine to get the blood sugar value. But this process is not only ...

AI Technical Summary

Problems solved by technology

In the existing food recognition, CNN is generally used for target detection and food classification, but in complex environments such as dim light, it is difficult to extract effective features or the extracted features are not prominent enough, thus affecting the accuracy of classification

Images