Image Based Food Calories Estimation Using Various Models of Machine Learning
Haoyu Hu, Zihao Zhang, Yulin Song
Abstract
For the last few decades, it has been the popular trend in China that people are putting more attention on improving their healthiness and regulating calorie intake for every meal, so that we build a model for calorie estimation of Chinese food. In an attempt to express our concerns on this issue, and with our great interests, we used object detection to estimate the calories count of some famous Chinese dishes as well as that of Western dishes. Based on the food images and previously defined calorie data, we built some image-based calorie estimation models, which we hoped can accurately identify the name of the Chinese and Western foods and provide their calorie intake and recipe, and finally offer meal plan advice for different groups of people. To identify the dish, we used the SSD (Single Shot MultiBox Detector) for real-time processing of object detection and classification. We also used a computer application called “labelImg” to manually label our dishes with their respective original dish names. Using our models, users can easily calculate the calorie intake of their desired foods by taking photos, saving a lot of time compared to their conventional methods.