30VNFoods: A Dataset for Vietnamese Foods Recognition

Abstract

This paper introduces a large dataset of 25136 images of 30 popular Vietnamese foods. Several machine learning and deep learning image classification techniques have been applied to test the dataset and the results were compared and report. A decent accuracy of 77.54% and a high top 5-accuracy of 96.07% were achieved. The dataset and the performance comparison of state-of-the-art algorithm tested on the dataset will be useful for ones to develop new food image classification algorithms.