Abstract: | People segmentation in images is very difficult due to the variability of different conditions, such as the position they adopt, background color, etc. To perform this segmentation, different techniques exist that, given an input image, they return a labelling of the different objects present in it. The purpose of this project is to conduct a comparison of recent techniques that allow multilabeling, and which are semi-automatic, specifically in the case of people segmentation. From an initial labeling identical for all the used methods, an analysis of them has been performed, evaluated over a public image dataset, and analyzing 2 points: interaction level, and efficiency. |