EgoBody: Human Body Shape and Motion of Interacting People from Head-Mounted Devices

Decoupling Human and Camera Motion from Videos in the Wild

“…We use only 3DPW to perform ablations on the reconstructed local pose using our method. The most relevant dataset that is captured with dynamic cameras and provides ground truth 3D pose in the global frame is the recently introduced EgoBody dataset [71]. Ego-Body is captured with a head-mounted camera on an interactor, who sees and interacts with a second interactee.…”

“…World PA First -MPJPE (W-MPJPE) reports the MPJPE after aligning the first frames of the prediction and the ground Table 3. Comparison with the state of the art on EgoBody dataset [71]. We compare our approach with a variety of stateof-the-art methods for human mesh recovery.…”

“…Our method acquires these global trajectories from videos in the wild, with no constraints on the capture setup, camera motion, or prior knowledge of the environment. Being able to do this from dynamic cameras is particularly relevant with the emergence of large-scale egocentric video datasets [5,9,71].…”

“…We evaluate our approach on EgoBody [71], a new dataset of videos captured with a dynamic (ego-centric) camera with ground truth 3D global human motion trajectory. Our approach achieves significant improvement upon the state-of-the-art method that also tries to recover the human motion without considering the signal provided by the background pixels [65].…”

Scene-Aware 3D Multi-Human Motion Capture from a Single Camera

“…Several current methods focus on positioning humans in a pre-scanned 3D scene [12,15,18] and on simultaneous estimation of human poses and objects humans interact with [7,59,62]. A different setup assumes an RGB-D sensor [68] or a moving camera [16,25,29,67] that facilitates estimating the scene geometry. Recent methods integrate physics-based constraints into monocular 3D human motion capture and mitigate foot-floor penetration and other severe artefacts [47,48].…”