GarSeM

GarSeM aims for a deep semantic understanding of garments which depict complex topological and physical behaviors, involving in particular strong self-occlusions or deformations of fabrics, making the problem very different from the traditional use of topological maps in scene understanding with rigid objects or pose estimation of articulated objects such as humans. The implementation of a vision-based deep semantic understanding of garments holds significant potential benefits beyond the aforementioned assistive tasks. These include promoting sustainable fashion through automated garment recycling, enhancing quality control and service in apparel shops, and more. Methods and techniques developed within the project to detect, segment and track soft and flexible objects with strong deformations and their parts could extend their impact beyond garment manipulation. For example, they could be applied to tasks dealing with organs in the medical field or plants in various agricultural/forestry-related scenarios.