Sensing and Estimating Human Pose for a Virtual Try-on Application

Virtual Try-on stands as a pivotal advancement in the clothing industry, offering a multitude of advantages for both businesses and consumers alike. In the sphere of Virtual Try-on, a myriad of algorithmic facets come into play, with Image-based (2D) Virtual Try-on recognized as a foundational stepping stone for those engaged in research and foundational model development. This paper remains firmly rooted in the realm of Image-based (2D) Virtual Try-on, with a primary focus on key algorithms, notably human parsing and human pose. Furthermore, our exploration extends to encompass other interrelated algorithms, encompassing body part segmentation and clothing keypoints detection, essential components of the Virtual Try-on process. The ultimate outcome of this research is the ability to seamlessly apply Virtual Try-on techniques to models donning well-fitted upper body clothing, in alignment with the specific scope of this paper.