Convolutional Neural Network and Dropout Technique for Recognition of Thai Food Image

Abstract

One of the main causes of diabetes, heart disease, and obesity in today's world is consumption behavior. To help people prevent such diseases, the consumption advice is required for behavior changes. A comparative study was proposed in this paper to find out a method for recognition of Thai food image using Convolutional Neural Network (CNN) and dropout technique to recognize the type of Thai food. CNN is used for distinguishing the complexity among Thai food images, while dropout technique is used for noise reduction. Sample used in the experiment were 16,170 images for recognizing 52 food types. The proposed method is compared with CNN without dropout, and Nu-In Net 1.1. The comparison results showed that CNN with dropout performed the best with 96.67% accuracy.