justmaiklas / seefood

Calorie Estimation from Images - bbm406 Fundamentals of Machine Learning Term Project

SeeFood

In this project, my goup mate and I built Food Calories Estimation System using with machine learning methods.

Course Web Page: BBM 406 Fall 2018 Fundamentals of Machine Learning

Our Blog Posts:

You can see what we did week by week with more details.

• https://medium.com/bbm406f18/week-1-seefood-be1097c7876a
• https://medium.com/bbm406f18/week-2-seefood-ae381ea34757
• https://medium.com/bbm406f18/week-3-seefood-a511dd2f17a7
• https://medium.com/bbm406f18/week-4-seefood-59f1b759b173
• https://medium.com/bbm406f18/week-5-seefood-f495a76ded70
• https://medium.com/bbm406f18/week-6-seefood-52720a73823d
• https://medium.com/bbm406f18/week-7-seefood-959bc06ec32

DataSet

We used ECUST Food Dataset. You can see more information about our used dataset ECUSTFD.

Object Detection

We used Faster R-CNN model to detect foods. We compare few different models then we obtained best results at Facter R-CNN inception v2 coco.

Volume Estimation & Calorie Estimation

Our baseline project used mathmatics formulas. We thought we can do better. Then we use Machine Learning Algorithms to estimate calories. We used Random Forest and K-Nearest Neighbors methods when we calculate calories.

Result

Method Name	Volume RMSE	Calorie RMSE
KNN	21.06	45.69
Random Forest	13.21	30.37

You can see more result and information about our project in our final report