Weakly-Supervised Mesh-Convolutional Hand Reconstruction in the Wild

Grasping Field: Learning Implicit Representations for Human Grasps

“…Recently, instead of estimating the hand skeleton, recovering the pose and the surface of the hand has become popular using statistical hand models, e.g., the MANO model [74], that can represent a variety of hand shapes and poses [25,93,97]. Using the template derived from MANO, [41] show that it is also possible to regress hand meshes directly using mesh convolution. In this work, we represent the 3D hand by a signed distance field, instead of a parametric hand model, due to the difficulty of incorporating object interaction into the model parameter space.…”

A Skeleton-Driven Neural Occupancy Representation for Articulated Hands

“…Hand pose estimation is a long-standing question and several learning-based approaches have been introduced. These approaches generally involve predicting 3D keypoint locations [8,13,16,18,19,28,40,41,43,54,57,58,60,64,66,71], regressing MANO [51] parameters [1,2,4,25,26,68], or directly predicting the full dense surface of the hand [20,30,39,61]. The methods that directly predict 3D key points usually achieve better performance, however, they do not yield dense surface which is crucial for hand interaction.…”

Capturing Implicit Spatial Cues for Monocular 3d Hand Reconstruction

“…[10] reconstructs the hand mesh by using graph convolutions in the spectral domain. [9] directly recover the mesh in image coordinates using a spatial mesh convolution. [11] propose an "image-to-lixel" prediction network.…”

“…approaches [9,10,11] that directly estimate hundreds of vertices to recover the hand mesh, model-based approaches are easier to learn, enjoying faster training and inference speed, which are more favored in real-world application [6]. However, it is highly no-linear to learn hand model parameters from a 2D image.…”