TRAM: Global Trajectory and Motion of 3D Humans from in-the-wild Videos

We propose TRAM, a two-stage method to reconstruct a human’s globaltrajectory and motion from in-the-wild videos. TRAM robustifies SLAM to recoverthe camera motion in the presence of dynamic humans and uses the scenebackground to derive the motion scale. Using the recovered camera as ametric-scale reference frame, we introduce a video transformer model (VIMO) toregress the kinematic body motion of a human. By composing the two motions, weachieve accurate recovery of 3D humans in the world space, reducing globalmotion errors by a large margin from prior work.https://yufu-wang.github.io/tram4d/

Further reading#

• Access Paper in arXiv.org