PEOPLECAP 2018

What is PeopleCap?

Accurately tracking, reconstructing, capturing and animating the human body in 3D is critical for human-computer interaction, games, special effects and virtual reality. In the past, this has required extensive manual animation.

Nowadays research in this area is allowing us to capture and learn realistic models of people and hands from real measurements coming from scans, depth cameras, color cameras and inertial sensors. Such a model is, ultimately, a compact parameterization of surface geometry that can be deformed to generalize to novel poses and shapes. This model can then be used to track bodies and hands from noisy sensors by optimizing the model parameters as to best fit noisy and incomplete image observations.

The workshop is intended to offer a meeting and discussion platform for researchers with diverse backgrounds, such as computer graphics, computer vision and optimization, and machine learning. This will hopefully push the state-of-the-art in “Capturing and modeling humans” in terms of models, methods and datasets.

The call for papers will be in the areas of

3D Human pose and shape estimation from images, depth cameras or inertial sensors
3D Hand pose estimation and tracking
Human body, hand and face modeling
3D/4D Performance capture of bodies, faces and hands
Capture of people and clothing
Human body and hand models
Models of human soft-tissue
Registration of bodies, hands and faces

While the computer vision community has seen a lot of work on methods for detecting and tracking people in 2D much less work has focused in reasoning directly in 3D. Hence, in PeopleCap, special emphasis will be given to methods that work in 3D and to methods that use a generative model.

Submission

Submission Deadline	July 24
Reviews Due	August 3
Notification of Acceptance	August 7
Camera-Ready Submission	August 17
Workshop	September 14

- Note: camera ready submission is now open.
- Note: submission deadline has been extended to July 24th.
- All deadlines are 5 PM Pacific time.
- Paper submissions should follow the exact same guidelines of ECCV, 14 pages plus references.
- Submissions can be uploaded to the CMT: https://cmt3.research.microsoft.com/PeopleCap2018/

If you do not have one already, create an account for cmt3 and login.
In case you are not directed to PeopleCap submission, type PeopleCap in the search box to find it.
Create a new submission and upload the main paper (and supplementary material if any).

- Accepted papers will be published in the proceedings of ECCV workshops.

Invited Speakers

Lourdes Agapito, University College London

Yaser Sheikh, Carnegie Mellon University

Adrian Hilton, University of Surrey

Jamie Shotton, Microsoft

Franziska Mueller, MPI for Informatics, Saarbrücken

Stefanie Wuhrer, INRIA

Program

13:30	Welcome and introduction
13:40	Weak Supervision for 3D Human Pose Estimation Lourdes Agapito
14:20	4D Performance Capture in the Wild Adrian Hilton
15:00	Towards Real-time Hand Tracking from In-the-wild Video" Franziska Mueller
15:40	Poster session and coffee break
16:40	Photorealistic Telepresence Yaser Sheikh
17:10	Generative Models of 3D Human Faces Stefanie Wuhrer
17:40	Closing remarks and best paper announcement

Accepted papers:

Nikolay N Chinaev (VisionLabs)*; Chigorin Alexander (VisionLabs); Ivan Laptev (INRIA Paris)
MobileFace: 3D Face Reconstruction with Efficient CNN Regression

Bastian Wandt (Leibniz University Hannover)*; Hanno Ackermann (Leibniz University Hannover); Bodo Rosenhahn (Leibniz University Hannover)
A Kinematic Chain Space for Monocular Motion Capture

Hang Dai (University of York)*; Nick Pears (University of York, UK); William Smith (University of York)
Non-rigid 3D Shape Registration using an Adaptive Template

Aaron Jackson (University of Nottingham)*; Chris Manafas (2B3D); Georgios Tzimiropoulos (University of Nottingham
3D Human Body Reconstruction from a Single Image via Volumetric Regression

Dylan Drover (Amazon Lab126)*; Rohith MV (Amazon Lab126); Ching-Hang Chen (Amazon Inc.); Ambrish Tyagi (Amazon Lab126); Amit Agrawal (Amazon Lab126); Cong Phuoc Huynh (Amazon)
Can 3D Pose be Learned from 2D Projections Alone?