1st International Workshop
on Understanding Human Activities
through 3D Sensors (UHA3DS'15)

May 8, 2015 Ljubljana, Slovenia
hosted by 11th IEEE International Conference on Automatic Face and Gesture Recognition FG 2015

Workshop Focus & Aims

Understanding human activities is a real problem, which needs an accurate acquisition of the movement sequence, consistent geometric representation of kinematics, dynamic modelling, and suitable learning process for motion identification. This workshop aims to bring together researchers from computer vision and machine learning communities, working together in a natural synergy and having an interest in using recent computing technologies to understand human, but also support him. In this way, we intend to present the recent vision-based algorithms in the related fields of static and temporal 3D data capture, modelling and representation, and their applications for social interactions. All aspects of 3D human sensing, such as detecting, tracking, motion and activity understanding will be addressed in this workshop. The covered topics include 3D pose estimation, human activity analysis, hand gesture analysis, body expression and body language. The workshop aims to provide an interactive platform for researchers to disseminate their most recent research results, discuss rigorously and systematically potential solutions and challenges, and promote new collaborations among researchers. In particular, the workshop is intended to target the following issues:

  • Accuracy of data acquisition from 3D sensors as a main challenge in action/activity recognition
  • Dynamic modelling of the movements
  • Learning statistical models of appearance and motion from a collection of videos
  • 3D joint/skeleton representation versus depth image representation of the action and limitations of each approach
  • Action recognition for humans-computer interactions
  • Activity recognition in presence of human-object interaction
  • Group activity recognition
  • Recognition of complex activities
  • Motion segmentation and representation of sub-actions and their contribution in recognition performance