The PIMPE project (pronounced /pim’pe/) is a project funded by the International Strategic Partnerships call of the “Investissements d’avenir” program (ANR-15-IDEX-02). The project funds a PhD student, Wen Guo, through a bilateral commitment between Université Grenoble Alpes and Institut de Robòtica Industrial of Universitat Politècnica de Catalunya. Wen is therefore co-advised between Dr. Francesc Moreno-Noguer and myself.
Estimating full human body pose is paramount in computer vision, and has a wide range of applicative scenarios, from the entertainment industry, to sports technology going through medical diagnosis. Current state-of-the-art addresses the multi-person pose estimation problem as multiple instances of single-person pose estimation and treats them independently of each other. However, multi-person interactive scenarios are characterized by coupled effects, strongly suggesting that the estimation of the pose of two (or more) bodies involved in an interaction should not be done independently. We identify a lack of methodological approaches for visual human estimation pose in multi-person interactive scenarios, and hypothesize is partly due to the lack of large-scale annotated training sets.
In this project we aim to develop machine learning techniques able to jointly estimate the full body pose of several persons involved in the same physical interaction. We are interested in a plethora of applicative scenarios, from industrial worker cooperation, to sports and arts as well as anomaly detection. To overcome the data availability issue, we propose to design learning strategies for the generation of realistic and controllable multi-person full-body extreme pose datasets: realistic so as to minimize the data distribution gap between training and real scenarios, controllable so as to generate a wide variability of poses, thus allowing for generalization to complex body motions and inter-body physical interactions.