Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Human Body Model Fitting by Learned Gradient Descent

About

We propose a novel algorithm for the fitting of 3D human shape to images. Combining the accuracy and refinement capabilities of iterative gradient-based optimization techniques with the robustness of deep neural networks, we propose a gradient descent algorithm that leverages a neural network to predict the parameter update rule for each iteration. This per-parameter and state-aware update guides the optimizer towards a good solution in very few steps, converging in typically few steps. During training our approach only requires MoCap data of human poses, parametrized via SMPL. From this data the network learns a subspace of valid poses and shapes in which optimization is performed much more efficiently. The approach does not require any hard to acquire image-to-3D correspondences. At test time we only optimize the 2D joint re-projection error without the need for any further priors or regularization terms. We show empirically that this algorithm is fast (avg. 120ms convergence), robust to initialization and dataset, and achieves state-of-the-art results on public evaluation datasets including the challenging 3DPW in-the-wild benchmark (improvement over SMPLify 45%) and also approaches using image-to-3D correspondences

Jie Song, Xu Chen, Otmar Hilliges• 2020

Related benchmarks

TaskDatasetResultRank
3D Human Pose EstimationHuman3.6M (test)--
547
3D Human Pose Estimation3DPW (test)
PA-MPJPE55.9
505
3D Human Pose EstimationHuman3.6M--
160
3D Human Pose and Shape Estimation3DPW (test)
MPJPE-PA56.4
158
3D Human Pose EstimationHuman3.6M Protocol #2 (test)--
140
Human Mesh Recovery3DPW--
123
3D Human Pose EstimationHuman3.6M (S9, S11)--
94
3D Human Pose and Shape Estimation3DPW
PA-MPJPE55.9
74
Human Mesh RecoveryHuman3.6M
Reconstruction Error56.4
47
3D Human Mesh Estimation3DPW (test)
PA-MPJPE55.9
44
Showing 10 of 16 rows

Other info

Follow for update