Image-based food recognition with a Conditional U-net

This project is about the food recognition challenge proposed by AICrowd, which consists in an image segmentation problem related to recognize food items in images. Recognizing food from images is a very useful tool for a variety of cases like sports, diets or medical issues, especially if people are allowed to track their daily food by just taking a photo with their mobile phones. Image-based food recognition is a very challenging topic because the same kind of food can look very different depending by how it is cooked, cut or other variables. Despite in the past few years a lot of progress were made, we still have problems about identification of food categories and ingredients. In this project is proposed an end-to-end approach through the use of a Conditional U-Net.