Food Calorie and Nutrition Analysis System based on Mask R-CNN

Meng Lin Chiang, Chia An Wu, Jian Kai Feng, Chiung Yao Fang, Sei-Wang Chen

研究成果: 書貢獻/報告類型會議貢獻

摘要

Over the past few decades, obesity has become a serious problem. Obesity is associated with many of the leading causes of death, such as chronic diseases including diabetes, heart disease, stroke, and cancer. The most effective way to prevent obesity is through food intake control, which involves understanding food ingestion, including the nutrients and calories of each meal. To assist with this issue, this study develops a food calorie and nutrition system that can analyze the composition of a food based on a provided image. Further, we introduce a newly collected dataset, Ville Cafe, for food recognition. This dataset contains 16 categories with 35,842 images, including salad, fruit, toast, egg, sausage, chicken cutlet, bacon, French toast, omelet, hash browns, pancake, ham, patty, sandwich, French fries, and hamburger. The system is based on a Mask Region-based Convolutional Neural Network (R-CNN) with a union postprocessing, which modifies the extracted bounding boxes and masks, without the non-maximum suppression (NMS), to provide a better result in both analytics and visualization. The recognition accuracy for the combination of Ville Cafe and the Food-256 Datasets was 99.86%, and the intersection over union (IoU) was 97.17%. The food weight estimation experiment included eight classes: salad, fruit, toast, sausage, bacon, ham, patty, and French fries. These classes respectively had 40, 40, 44, 40, 41, 49, 26, and 40 data points, adding up to 320 data points for the linear regression model. In the experimental results, the average absolute error was 8.22, and the average relative error was 0.13.

原文英語
主出版物標題2019 IEEE 5th International Conference on Computer and Communications, ICCC 2019
發行者Institute of Electrical and Electronics Engineers Inc.
頁面1721-1728
頁數8
ISBN(電子)9781728147437
DOIs
出版狀態已發佈 - 2019 十二月
事件5th IEEE International Conference on Computer and Communications, ICCC 2019 - Chengdu, 中国
持續時間: 2019 十二月 62019 十二月 9

出版系列

名字2019 IEEE 5th International Conference on Computer and Communications, ICCC 2019

會議

會議5th IEEE International Conference on Computer and Communications, ICCC 2019
國家中国
城市Chengdu
期間19/12/619/12/9

ASJC Scopus subject areas

  • Information Systems and Management
  • Modelling and Simulation
  • Artificial Intelligence
  • Computer Networks and Communications
  • Computer Science Applications
  • Hardware and Architecture
  • Information Systems

指紋 深入研究「Food Calorie and Nutrition Analysis System based on Mask R-CNN」主題。共同形成了獨特的指紋。

引用此