Gravity-Aware Monocular 3D Human-Object Reconstruction

Dabral, Rishabh; Shimada, Shigetaka; Jain, Arjun; Theobalt, Christian; Golyanik, Vladislav

doi:10.1109/iccv48922.2021.01214

Cited by 31 publications

(19 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Yu et al [63] also support composite scenes in the parcours and sports scenarios. Although there is a growing interest in investigating the interactions of humans and objects [2,10], 3D motion capture of multiple humans with environmental awareness from a single monocular camera remains underexplored.…”

Section: Scene-aware Motion Capturementioning

confidence: 99%

“…Bieler et al [3] estimate the height of a single person from monocular videos by observing jumping people. Dabral et al [10] require an interaction with an object undergoing a free flight to resolve the absolute scene scale. Both methods assume motion influenced by the universal law of gravity near the surface of Earth, which allows them to relate the time spent in the air or the form of the observed trajectory with absolute distances in the metric units.…”

Section: Scene-aware Motion Capturementioning

confidence: 99%

“…Our method is implemented in PyTorch [38] using Py-Torch3D [43] for the rasterization ( 6) and silhouette rendering (10). The camera intrinsics are used in the 3D joint projection ( 9), rasterization (6), and rendering (10) parts, and can be obtained from video metadata if not given.…”

Section: Implementation Detailsmentioning

confidence: 99%

See 2 more Smart Citations

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

Luvizon¹,

Habermann²,

Golyanik³

et al. 2023

Preprint

View full text Add to dashboard Cite

Section: Scene-aware Motion Capturementioning

confidence: 99%

Section: Scene-aware Motion Capturementioning

confidence: 99%

See 1 more Smart Citation

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

Luvizon¹,

Habermann²,

Golyanik³

et al. 2023

Preprint

View full text Add to dashboard Cite

“…There are several methods that explore different kinds of HSI; these can be divided into three categories by the interaction granularity between the human and scene: (1) Hand-Object [8,9,34,49,54,81,91]. (2) Body-Object [16,43,53,80,96]. (3) Body-Scene [10,26,30,63,74,87].…”

Section: Related Workmentioning

confidence: 99%

Human-Aware Object Placement for Visual Environment Reconstruction

Yi¹,

Huang²,

Tzionas³

et al. 2022

Preprint

View full text Add to dashboard Cite

Figure 1. From a monocular video sequence, MOVER reconstructs a 3D scene that best affords humans interacting with it. Existing methods for monocular 3D scene reconstruction ignore people and produce physically implausible scenes. MOVER takes as input: (1) several images of human-scene interaction (HSI) from a static camera, (2) a rough estimate of 3D object shape and placement in 3D space [64], and (3) estimated 3D human bodies interacting with the scene [45,66]. Each frame contains valuable information about humans, objects, and the proximal relationship between them. MOVER accumulates this information across frames, to optimize for a physically plausible 3D scene. The final 3D scene is more accurate than the input and enables reasoning about human-scene contact.

show abstract

“…A few methods exploit various scene constraints during the optimization process to improve depth prediction [95,106]. Alternatively, recent approaches use physics-based constraints to ensure the physical plausibility of the estimated poses [12,34,84,98,104]. Iqbal et al [32] exploit a limblength constraint to recover the absolute translation of the person using a 2.5D representation.…”

Section: Related Workmentioning

confidence: 99%