【講題】Bridging Realism and Animation: 3D Human Pose and Shape Estimation and Beyond

【演講時間】6月5日(星期五)下午1點30分  

【演講地點】清華大學校本部第二綜合大樓B側8樓A813室

【摘要】

Creating realistic and controllable 3D human motion is essential for animation, film production, virtual concerts, and immersive media. However, traditional motion capture systems and manual keyframing are often costly, labor-intensive, and difficult to scale. This talk presents a series of data-driven approaches that bridge real-world visual observations and animation-oriented 3D content creation. I will first introduce our works on monocular video-based 3D human pose and shape estimation, focusing on temporal modeling and multi-candidate motion modeling for accurate and stable reconstruction. I will then discuss extensions beyond estimation, including 3D human motion in-betweening, camera trajectory authoring for virtual concerts, and part-aware text-to-motion generation. Together, these works aim to make 3D motion and camera authoring more accessible, controllable, and suitable for next-generation digital content creation.