|Jason Y. Zhang* 1||Sam Pepose* 2||Hanbyul Joo 2|
|Deva Ramanan 1,3||Jitendra Malik 2,4||Angjoo Kanazawa 4|
|1 Carnegie Mellon University||2 Facebook AI Research|
|3 Argo AI||4 University of California, Berkeley|
|In ECCV 2020|
We present PHOSA: Perceiving Human-Object Spatial Arrangements. From a single RGB image, we recover plausible spatial arrangements of humans and objects by reasoning about their intrinsic scale and human-object interaction. Here, we demonstrate our approach on images captured in unconstrained outdoor environments across a wide range of object categories.
Perceiving 3D Human-Object Spatial Arrangements from a Single Image in the Wild
Jason Y. Zhang*, Sam Pepose*, Hanbyul Joo, Deva Ramanan, Jitendra Malik, and Angjoo Kanazawa