The Calorie Estimation Project can be mainly divided into two parts, identifying food from image, and estimating calorie from certain food image. For the identification of food image, we performed multi-class SVM algorithms, with different features explored and compared, including HOG (Histogram of Gradients), LBP (Linear Binary Pattern) and CNN. The result shows that the local feature LBP performs the best overall. The food calorie data from Internet is collected to conclude a table for easy conversion from food category to calorie.
The Calorie Estimation Project can be mainly divided into two parts, identifying food from image, and estimating calorie from certain food image. For the identification of food image, we performed multi-class SVM algorithms, with different features explored and compared, including HOG (Histogram of Gradients), LBP (Linear Binary Pattern) and CNN. The result shows that the local feature LBP performs the best overall. The food calorie data from Internet is collected to conclude a table for easy conversion from food category to calorie.