Human body segmentation into particular body parts (regions) is a crucial task in many human body-oriented computer vision applications. In general, body segmentation as a preprocessing step can be helpful in understanding the skeletal and body structure of a human subject. Applying the segmentation methods on 3D human data is an important task, since 3D data has potential to provide additional information over RGB images, and tends to yield more accurate results.
The aim of the bachelor thesis is to examine the human body segmentation task and apply various methods on human body datasets for the purpose of body part segmentation into individual regions with the highest possible accuracy. Evaluate and compare these methods and sumarize the results. Part of the work is to to study the principles of neural networks and appropriate deep learning framework. The thesis focuses on researching exclusively synthetic data.