Data set labeling method and system

Abstract

The invention discloses a data set labeling method, which comprises the following steps of: training a preset food material training set according to a MobileNet neural network, and constructing a food material recognition MobileNet initial model; according to the food material identification MobileNet initial model, carrying out food material identification on a plurality of dish pictures, and obtaining food material information in each dish picture; and according to the food material information, marking a classification label corresponding to each dish picture. The data set marking efficiency and accuracy are improved, and the workload of manual marking is greatly reduced.

Description

technical field [0001] The invention belongs to the technical field of identification, and in particular relates to a data set labeling method and system. Background technique [0002] In the field of recognition technology, it is usually necessary to label sample data first, and then use the labeled sample data to learn and train a neural network model to obtain a neural network model for recognition. [0003] In order to obtain labeled sample image data for learning and training, most of the existing methods use manual labeling to label the collected sample data. It is difficult to label the dataset used for ingredient recognition neural network training, because it is difficult to use a search engine to search for all the ingredients in a dish picture through keyword searches. For a neural network that recognizes 100 ingredients, about 1 million pictures of dishes are needed for training. If the ingredients are marked one by one on 1 million pictures of dishes, assuming...

AI Technical Summary

Problems solved by technology

Therefore, the work of manual labeling is cumbersome and repetitive, the labeling speed is slow, the efficiency is low, and the accuracy is poor

Images