Street Food Detection of Bangladesh Using Deep Learning

Abstract

Street food demand is increasing day by day in Bangladesh. It has yummy tastes, easily accessibility, low price, easily made, easy to available, attraction to the foods, and above all, needs of the street people. There are many different classes of people from many different areas, especially the middle class, poor and the lower-class people come in Dhaka in search job for better earning. And their earning is so low that’s why they can buy this type of street food because of low price. Mainly most of the young people eating foods at the street and it is a fashion nowadays. This study has been detected of these street foods can help people detect them. To conduct this study, we used Deep Learning process to build our system of recognition various street cuisine. Deep learning is a strong technology that has been used in a variety of fields to automate fundamental procedures and improve the outcomes of these operations. A total of 3023 images with 14 items of street foods were used to detect. We conduct this captured image and gained our expected feature by using image classification. For image classification of street foods, we used TensorFlow algorithm. We also used Convolutional Neural Network (CNN) for architecture and feature extraction of our Model. The Convolutional Neural Network (CNN) achieved the accuracy of 97.72% , which is good and also giving us inspiration for our next research. Deep learning is being utilized on the field and in the marketplace to boost yield and ensure the quality of street food reaching consumers in the area of street food detection. In this thesis, we planned to create a simple CNN that can recognize street food in images. This system would help the human to reduce the time and effort needed for detecting of street foods at street. We used sequential model to build our system. Deep Learning (DL) has several applications due to its ability to learn robust representations from images. Convolutional Neural Networks (CNN) is the main DL architecture for image classification.