SWVie-Food: A Dataset for Recognizing Foods in Southwest Vietnam Based on Deep Learning

• Published in: 2022 RIVF International Conference on Computing and Communication Technologies (RIVF) pp. 488 - 493
• Main Authors: Lam, Khang Nhut, Nguyen, My-Khanh Thi, Nguyen, Khang Duy, Nguyen, Nghia Hieu, Nguyen, Kim-Yen Thi, Ware, Andrew
• Format: Conference Proceeding
• Language: English
• Published: IEEE 20-12-2022
• Subjects: Biological system modeling; Cellular phones; Deep learning; Economics; food recognition; Image recognition; MobileNet; Real-time systems; SWVie-food; Transformers; Vision Transformer; YOLO
• DOI: 10.1109/RIVF55975.2022.10013882

Recognizing foods in general and foods in Vietnam, in particular, has been the subject of many studies. However, we have found no attempts to recognize popular local foods from the southwest of Vietnam. In this study, we introduce an SWVie-Food dataset consisting of 3,022 images from 50 categories of local foods in southwest Vietnam. Then, we fine-tune deep learning models for recognizing foods. The vision Transformer model for food recognition achieves a mAP of 0.931, which outperforms MobileNetV3, YOLOv5, and YOLOv7 with mAP scores of 0.772, 0.831, and 0.870, respectively. Finally, we build a mobile application to assist users to recognize foods from uploaded images or real-time. In addition, our application provides information about the origin, ingredients, and recipes of local foods. Our study contributes to advertising tourism in the southwest of Vietnam.