This paper presents a learning-based approach to perform human shape transfer between an arbitrary 3D identity mesh and temporal motion sequence of meshes. Recent approaches tackle the pose on per-frame basis do not yet consider valuable information about dynamics, e.g., body or clothing inherently present in sequences. datasets provide such sequences meshes, this work investigates how leverage associated intrinsic features order improve transfer. These are expected help preserve consistency...

This paper presents a hybrid skeleton-driven surface registration (HSDSR) approach to generate temporally consistent meshes from multiple view video of human subjects. 2D pose detections are used estimate 3D skeletal on per-frame basis. The is embedded into reconstruction allowing any frame be reposed the shape other in captured sequence. Skeletal motion transfer performed by selecting reference data and reposing it match estimation frames allows an initial coarse alignment prior refinement...

This paper introduces Deep4D a compact generative representation of shape and appearance from captured 4D volumetric video sequences people. achieves highly realistic reproduction, replay free-viewpoint rendering actor performance multiple view acquisition systems. A deep network is trained on an performing motions to learn model the dynamic appearance. We demonstrate proposed can provide encoded capable high-quality synthesis with two orders magnitude compression. variational...

This paper presents techniques to animate realistic human-like motion using a compressed learnt model from 4D volumetric performance capture data. Sequences of dynamic geometry representing human performing an arbitrary are encoded through generative network into compact space representation, whilst maintaining the original properties, such as, surface dynamics. An animation framework is proposed which computes optimal graph novel capabilities compression and synthesis properties network....

To address the need for interoperable user representations (avatars) cross-platform exchange format immersive realities, ISO/IEC JTC 1/SC29/WG03 MPEG Systems has standardized a Scene Description framework in 23090-14 [ISO/IEC 2023]. It serves as baseline representation to enrich interactive experience between 3D objects an scene. This work presents Original Reference Geometric Avatar Neutral (Morgan), humanoid avatar specified informative content MPEG-I (MPEG-I SD) standardization group....

We present an unsupervised data-driven approach for non-rigid shape matching. Shape matching identifies correspondences between two shapes and is a fundamental step in many computer vision graphics applications. Our designed to be particularly robust when digitized using 3D scanners that contain fine geometric detail suffer from different types of noise including topological caused by the coalescence spatially close surface regions. build on strategies. First, hierarchical patch based...

We consider the problem of modifying/replacing shape style a real moving character with those an arbitrary static source character. Traditional solutions follow pose transfer strategy, from to shape, that relies on skeletal parametrization. In this paper, we explore alternative approach transfers onto The expected benefit is avoid inherently difficult conversion required parametrization applied characters. To purpose, image techniques and investigate how adapt them 3D human shapes. Adaptive...