A Convolution Neural Network Based Classification Approach for Recognizing Traditional Foods of Bangladesh from Food Images

Abstract

Research has become one of the most talked buzzwords in the modern era due to knowledge development and practical improvement obtained by research papers. However, choosing the exact research topic from the vast ocean of knowledge fields for an unaware individual person is quite hard. Sometimes a beginner researcher/student cannot specify the research interest, or they are unfamiliar with trending topic and technology, which are most impactful in the near future. Therefore, we think computer vision is such an important thing that already helps the human being in a various way and it will help us in the future. The process of identifying food items from an image is one of the promising applications of visual object recognition in computer vision. However, analysis of food items is a particularly challenging work due to the characteristic of their images, which is why a low classification accuracy has achieved by traditional approaches in the field. Deep neural networks have exceeded such solutions. With a goal to successfully applying computer vision techniques to classify food images founded on Inception-v3 model of TensorFlow platform, we conduct the transfer learning technique to retrain the food category datasets. Our methodology demonstrates auspicious outcomes with an average precision of nearly 95.2% in properly identifying among seven traditional Bangladeshi foods. This research paper intends to give details about various classification techniques and process by using computer vision methods that are being used in today‟s research for classifying food items.