High-fidelity pixel-aligned image-to-3D generation.
Detect and visualize human poses with 308 keypoints