AvatarCap: Animatable Avatar Conditioned Monocular Human Volumetric Capture

MoRF: Mobile Realistic Fullbody Avatars from a Monocular Video

“…Modeling geometry of dynamic non-rigid scenes is considered in a number of recent approaches either for capturing human actors [5,54,29,27,30,35,16,20,17,59] or general scenes [32,45,41]. Towards this end, 3D scans are required for supervision in [20,17,35,54,29] to learn rigged human geometry.…”

“…Modeling geometry of dynamic non-rigid scenes is considered in a number of recent approaches either for capturing human actors [5,54,29,27,30,35,16,20,17,59] or general scenes [32,45,41]. Towards this end, 3D scans are required for supervision in [20,17,35,54,29] to learn rigged human geometry. Likewise, [32,30,16,5,48,27] utilize multi-view data to capture appearance and produce photo-realistic avatars under arbitrary viewpoints and in arbitrary poses.…”

“…Our neural texture warping allows for consistent learning of the neural texture from different poses and views without training on synthetic data [64], excluding most frames from texture learning [47], or over-fitting the neural renderer to unmodeled variations [15]. Many of the mentioned works introduce learned submodules [32,16,30,54,29,27,64] or pipeline stages [2,3] to estimate additional geometry warping in 3D canonical (unposed or view-agnostic) space. Likewise, [54,29,27,30] parametrize the 3D warping fields or displacements [35,64] as neural networks conditioned on human pose to facilitate learning of complex pose-dependent geometry.…”

“…Many of the mentioned works introduce learned submodules [32,16,30,54,29,27,64] or pipeline stages [2,3] to estimate additional geometry warping in 3D canonical (unposed or view-agnostic) space. Likewise, [54,29,27,30] parametrize the 3D warping fields or displacements [35,64] as neural networks conditioned on human pose to facilitate learning of complex pose-dependent geometry. Methods for general scenes leverage other variants of conditioning.…”