Malaysian Food Recognition using Alexnet CNN and Transfer Learning

This paper focuses its study on food recognition using one type of deep learning to support calorie estimation among variation of Malaysian food. Previously available studies of food calorie estimation have not yet discussed the Malaysian diet and its calories specifically. Therefore, this will be the focus of this study where datasets are built around the Malaysian diet. The method used in this study is image recognition based on convolution neural network called Alexnet. In addition to Alexnet CNN, a method called Transfer Learning is also used where customised dataset based on Malaysian food is put together and will later be applied to Transfer Learning. The resulting computer programme has produced a high accuracy of 91.43% due to the vast network of the Alexnet. This will be useful as a start to food calorie estimation studies based on the Malaysian diet.