MotionCrafter: Dense Geometry and Motion Reconstruction with a 4D VAE

1. 1Jointly reconstruct 4D geometry and estimate dense motion from a monocular video
2. 2Represent dense 3D point maps and 3D scene flows in a shared coordinate system
3. 3Utilize a 4D VAE to learn the joint representation effectively
4. 4Apply the learned representation to various applications such as robotics, autonomous driving, and animation

Computer vision engineers and researchers benefit from this framework as it enables dense geometry and motion reconstruction, which can be applied to various applications such as robotics, autonomous driving, and animation

💡 MotionCrafter leverages a 4D VAE to jointly reconstruct 4D geometry and estimate dense motion from a monocular video

💡 MotionCrafter: 4D geometry & motion reconstruction from monocular video using 4D VAE!