Perzeptive Systeme Optics and Sensing Laboratory Conference Paper 2022

Capturing and Inferring Dense Full-Body Human-Scene Contact

Screen shot 2022 05 28 at 1.55.51 pm

Inferring human-scene contact (HSC) is the first step toward understanding how humans interact with their surroundings. While detecting 2D human-object interaction (HOI) and reconstructing 3D human pose and shape (HPS) have enjoyed significant progress, reasoning about 3D human-scene contact from a single image is still challenging. Existing HSC detection methods consider only a few types of predefined contact, often reduce body and scene to a small number of primitives, and even overlook image evidence. To predict human-scene contact from a single image, we address the limitations above from both data and algorithmic perspectives. We capture a new dataset called RICH for “Real scenes, Interaction, Contact and Humans.” RICH contains multiview outdoor/indoor video sequences at 4K resolution, ground-truth 3D human bodies captured using markerless motion capture, 3D body scans, and high resolution 3D scene scans. A key feature of RICH is that it also contains accurate vertex-level contact labels on the body. Using RICH, we train a network that predicts dense body-scene contacts from a single RGB image. Our key insight is that regions in contact are always occluded so the network needs the ability to explore the whole image for evidence. We use a transformer to learn such non-local relationships and propose a new Body-Scene contact TRansfOrmer (BSTRO). Very few methods explore 3D contact; those that do focus on the feet only, detect foot contact as a post-processing step, or infer contact from body pose without looking at the scene. To our knowledge, BSTRO is the first method to directly estimate 3D body-scene contact from a single image. We demonstrate that BSTRO significantly outperforms the prior art. The code and dataset are available at https://rich.is.tue.mpg.de.

Author(s): Huang, Chun-Hao P and Yi, Hongwei and Höschle, Markus and Safroshkin, Matvey and Alexiadis, Tsvetelina and Polikovsky, Senya and Scharstein, Daniel and Black, Michael J.
Book Title: 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2022)
Pages: 13264--13275
Year: 2022
Month: June
Publisher: IEEE
Project(s):
Bibtex Type: Conference Paper (inproceedings)
Address: Piscataway, NJ
DOI: 10.1109/CVPR52688.2022.01292
Event Name: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2022)
Event Place: New Orleans, Louisiana
State: Published
URL: https://rich.is.tue.mpg.de/
Electronic Archiving: grant_archive
ISBN: 978-1-6654-6947-0
Links:

BibTex

@inproceedings{huang2022rich,
  title = {Capturing and Inferring Dense Full-Body Human-Scene Contact},
  booktitle = {2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2022)},
  abstract = {Inferring human-scene contact (HSC) is the first step toward understanding how humans interact with their surroundings. While detecting 2D human-object interaction (HOI) and reconstructing 3D human pose and shape (HPS) have enjoyed significant progress, reasoning about 3D human-scene contact from a single image is still challenging. Existing HSC detection methods consider only a few types of predefined contact, often reduce body and scene to a small number of primitives, and even overlook image evidence. To predict human-scene contact from a single image, we address the limitations above from both data and algorithmic perspectives. We capture a new dataset called RICH for “Real scenes, Interaction, Contact and Humans.” RICH contains multiview outdoor/indoor video sequences at 4K resolution, ground-truth 3D human bodies captured using markerless motion capture, 3D body scans, and high resolution 3D scene scans. A key feature of RICH is that it also contains accurate vertex-level contact labels on the body. Using RICH, we train a network that predicts dense body-scene contacts from a single RGB image. Our key insight is that regions in contact are always occluded so the network needs the ability to explore the whole image for evidence. We use a transformer to learn such non-local relationships and propose a new Body-Scene contact TRansfOrmer (BSTRO). Very few methods explore 3D contact; those that do focus on the feet only, detect foot contact as a post-processing step, or infer contact from body pose without looking at the scene. To our knowledge, BSTRO is the first method to directly estimate 3D body-scene contact from a single image. We demonstrate that BSTRO significantly outperforms the prior art. The code and dataset are available at https://rich.is.tue.mpg.de.},
  pages = {13264--13275},
  publisher = {IEEE},
  address = {Piscataway, NJ},
  month = jun,
  year = {2022},
  slug = {huang2022rich},
  author = {Huang, Chun-Hao P and Yi, Hongwei and H{\"o}schle, Markus and Safroshkin, Matvey and Alexiadis, Tsvetelina and Polikovsky, Senya and Scharstein, Daniel and Black, Michael J.},
  url = {https://rich.is.tue.mpg.de/},
  month_numeric = {6}
}