aical

I love this idea but the execution leaves a lot to be desired - I took a photo of my hand and it returned 105 calories. It worked on a papaya but as another commenter said, it doesn't return any classification about the detected food that would give me confidence that it correctly identified what I snapped. It really should return a detected food label. It also returned the same number of calories on a small papaya and a large papaya... I suspect you're using an off-the-shelf convolutional network trained on ImageNet combined with the USDA calories database or are you using a custom trained model?

Cool upp!!!