Food image automatic classification method based on convolutional neural networks

Abstract

The invention discloses a food image automatic classification method based on convolutional neural networks. The food image automatic classification method comprises steps that 1) food image data is crawled by a network crawler from Internet, and food images having correct labels are artificially screened out and filed in an Initial Data data set; 2) the network crawler is used to search a lot of food image data of the target classification in a mainstream search engine and an image sharing website, and at the same time, a step 4) is executed regularly; 4) a Food CNN network is used for data screening, and the data is divided into a Crawl Data and a Noisy Data; 5) the expanded data Crawl Data is used to update the Food CNN network; 6) whether the Noisy Data data size is reasonable is determined, and whether the crawling is continued is determined; 7) the crawling is stopped, and a Food Final CNN is trained. A correction rate is high.

Description

technical field [0001] The invention belongs to a method for automatically classifying food images based on a convolutional neural network, and relates to a convolutional neural network, web crawler technology and image classification technology. Background technique [0002] In recent years, with the improvement of people's quality of life, people's pursuit of food has become more and more diversified, but the variety of dishes is also dizzying. In addition, a healthy diet is gradually becoming one of the popular trends in people's pursuit of quality of life. The key to this is the acquisition of food information. How to obtain dish information through simple pictures or photos, such as calorie content, nutritional content, etc., is one of the important means to simplify food information acquisition, and automatic food identification is an important step in food information acquisition. [0003] Food images are changeable, complex and diverse. Traditional image recognitio...

AI Technical Summary

Problems solved by technology

[0008] In order to overcome the shortcomings of the food image data obtained by web crawlers in the prior art that the correct rate of the training classifier is low due to excessive data noise, the present invention proposes a method to effectively avoid the training classifier accuracy rate caused by excessive data noise. An automatic food image classification method based on convolutional neural network with high correct rate in low cases. The convolutional neural network algorithm directly uses the image as input, avoiding the complicated feature extraction and data reconstruction process in the traditional recognition algorithm. The classification accuracy and robustness of the classifier are high

Images