• 강사 : 김태균 교수 (KAIST 전산학부)
• 일시 : 2023. 4. 24 (월) 16:00~17:30

3D hand pose estimation plays a key role for novel man-machine interfaces. Estimating 3D postures, i.e. regressing 3D locations of all hand joints, given highly articulated and deformable hands, is challenging. The problem is also highly related to human body pose, and facial landmarking, yet has unique challenges. Dr. T-K Kim has tackled the problem by various novel ideas on top of cutting-edge techniques, pushing the envelope in the field. This talk briefly overviews the advances achieved by the computer vision and learning lab led by T-K. Kim, then presents two latest innovations. His team proposed a novel solution to estimate hand poses under hand-object interaction. It is achieved by removing objects from hand images via GAN and a hand mesh model (CVPR2020 best paper finalist). Most recently, his team developed an impressively accurate solution to reconstruct two interacting hands by novel implicit functions. The solution is resolution-free and offers precise alignments with input images (CVPR2023).

* 강연자료