MotioNet: 3D Human motion reconstruction from monocular video with skeleton consistency

被引：0

作者：

Shi, Mingyi ^{[1
]}

Aberman, Kfir ^{[3
]}

Aristidou, Andreas ^{[5
]}

Komura, Taku ^{[7
]}

Lischinski, Dani ^{[2
]}

Cohen-Or, Daniel ^{[4
]}

Chen, Baoquan ^{[6
]}

机构：

[1] Shandong University, Aicfve, Beijing, China

[2] Aicfve, Beijing Film Academy, Tel-Aviv University, Israel

[3] University of Cyprus and Rise Research Centre, Cyprus

[4] Edinburgh University, Japan

[5] Shandong University, China and the Hebrew University of Jerusalem, Israel and Aicfve, Israel

[6] Tel-Aviv University, Israel and Aicfve, Beijing Film Academy, Israel

[7] Cfcs, Peking University, Aicfve, Beijing, China

来源：

ACM Transactions on Graphics | 2020年 / 40卷 / 01期

基金：

以色列科学基金会; 欧盟地平线“2020”;

关键词：

Rotation - Deep neural networks - Image reconstruction - Musculoskeletal system - Inverse problems - Three dimensional computer graphics - Inverse kinematics;

D O I：

暂无

中图分类号：

学科分类号：

摘要：

We introduce MotioNet, a deep neural network that directly reconstructs the motion of a 3D human skeleton from a monocular video. While previous methods rely on either rigging or inverse kinematics (IK) to associate a consistent skeleton with temporally coherent joint rotations, our method is the first data-driven approach that directly outputs a kinematic skeleton, which is a complete, commonly used motion representation. At the crux of our approach lies a deep neural network with embedded kinematic priors, which decomposes sequences of 2D joint positions into two separate attributes: a single, symmetric skeleton encoded by bone lengths, and a sequence of 3D joint rotations associated with global root positions and foot contact labels. These attributes are fed into an integrated forward kinematics (FK) layer that outputs 3D positions, which are compared to a ground truth. In addition, an adversarial loss is applied to the velocities of the recovered rotations to ensure that they lie on the manifold of natural joint rotations. The key advantage of our approach is that it learns to infer natural joint rotations directly from the training data rather than assuming an underlying model, or inferring them from joint positions using a data-agnostic IK solver. We show that enforcing a single consistent skeleton along with temporally coherent joint rotations constrains the solution space, leading to a more robust handling of self-occlusions and depth ambiguities. © 2020 ACM.

引用