Project TrueSize

The project TrueSize aims the goal of body measurement with mobile devices. For example, videos of people are created using smartphones and, as a result, three-dimensional information is extracted. The 3D models created in this way can subsequently be used, for example, to improve clothing sizes, which are currently only updated very sporadically and, moreover, are not standardized. Through this development, users (producers, distributors, …) would have access to up-to-date measurement tables, could monitor and respond to global changes.

Duration: 01.05.2019 – 31.07.2021
Partner: TrueSize
Funding: Austrian Research Promotion Agency FFG – Basisprogramm

Publications 2

2021

Zwettler G., Praschl C., Baumgartner D., Zucali T., Turk D., Hanreich M., Schuler A. (2021). Three-step Alignment Approach for Fitting a Normalized Mask of a Person Rotating in A-Pose or T-Pose Essential for 3D Reconstruction based on 2D Images and CGI Derived Reference Target Pose.In Proceedings of the 16th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications – Volume 5: VISAPP, ISBN 978-989-758-488-6, pages 281-292. DOI: 10.5220/0010194102810292.

Visual Computing

DOI: 10.5220/0010194102810292

The 3D silhouette reconstruction of a human body rotating in front of a monocular camera system is a very challenging task due to elastic deformation and positional mismatch from body motion. Nevertheless, knowledge of the 3D body shape is a key information for precise determination of one’s clothing sizes, e.g. for precise shopping to reduce the number of return shipments in online retail. In this paper a novel three step alignment process is presented, utilizing As-Rigid-As-Possible (ARAP) transformations to normalize the body joint skeleton derived from OpenPose with a CGI rendered reference model in A- or T-pose. With further distance-map accelerated registration steps, positional mismatches and inaccuracies from the OpenPose joint estimation are compensated thus allowing for 3D silhouette reconstruction of a moving and elastic object without the need for sophisticated statistical shape models. Tests on both, artificial and real-world data, generally proof the practicability of th is approach with all three alignment/registration steps essential and adequate for 3D silhouette reconstruction data normalization.

2020

Baumgartner D., Praschl C., Zucali T., Zwettler G. Hybrid Approach for Orientation-Estimation of Rotating Humans in Video Frames Acquired by Stationary Monocular Camera.

Visual Computing Machine Learning

Accurate human orientation estimation with respect to the POSE of a monocular camera system is a challenging task due to general aspects of camera calibration and the deformability of a moving human body. Therefore, novel deep learning approaches for precise object position determination in robotics are difficult to adapt for human body analysis. In this work, we present a hybrid approach for accurately estimating a human body relative to a camera system, significantly improving the results derived from poseNet by applying optical flow analysis in a frame-to-frame comparison. The human body, which rotates in the T-position in situ, is thereby center-aligned, with object tracking methods applied to compensate for translations of the body motion. After 2D skeletal extraction, optical flow is calculated for an ROI region aligned relative to the vertical skeletal junction representing the spine and compared frame by frame. To evaluate the suitability of clothing as a basis for good features, local pixel homogeneity is considered to constrain optical flow to heterogeneous regions with distinguishing features such as imprint patterns, buttons, or buckles in addition to local illumination change. Based on the mean optical flow with rough approximation of the axial body shape as an ellipse, accuracy between 0.1° and 2.0° is achieved for orientation estimation on a frame-to-frame comparison evaluated and validated on both CGI renderings and real videos of people wearing clothes with different features.