Abstract: Food Recognition is a computer vision application that has gained huge research interest. Food recognition and classification system help to understand foods’ different diversity as different cultures entails different cuisines. There is need to recognize and document Nigerian Foods to avoid going into extinction. It is also necessary to understand the food specific estimated calorie contents. Existing food recognition systems are based on foods found in developed nations. This paper presents a real-time Android system for Nigerian food recognition and documentation system using finetuned MobileNetV2 model. The Convolutional Neural Network (CNN) is effective in image recognition and classification. Therefore, MobileNetV2, a CNN-based deep learning model, is utilized to recognize and classify 10 Nigerian food classes from 500 food images of our dataset developed from scratch known as the NaijaFood101. We evaluated the Nigerian food learning model’s performance to determine the accuracy of the food recognition and classification system. The model achieved an average of 97% recognition accuracy on the evaluation or test data.

Keywords: Food recognition, calorie estimation, NaijaFood101 dataset, Nigerian foods, deep learning, convolutional Neural Network

PDF | DOI: 10.17148/IJARCCE.2022.111201

Open chat
Chat with IJARCCE