Publications
A Hierarchy-Aware Pose Representation for Deep Character Animation
Nefeli Andreou, Andreas Lazarou, Andreas Aristidou, Yiorgos Chrysanthou
arXiv.org > cs > arXiv:2111.13907, Nov 2021
In this work we present an efficient method for training neural networks, specifically designed for character animation. We use dual quaternions as the mathematical framework, and we take advantage of the skeletal hierarchy, to avoid rotation discontinuities, a common problem when using Euler angle or exponential map parameterizations, or motion ambiguities, a common problem when using positional data. Our method does not requires re-projection onto skeleton constraints to avoid bone stretching violation and invalid configurations, while the network is propagated learning using both rotational and positional information.
Rhythm is a Dancer: Music-Driven Motion Synthesis with Global Structure
Andreas Aristidou, Anastasios Yiannakidis, Kfir Aberman, Daniel Cohen-Or, Ariel Shamir, Yiorgos Chrysanthou
arXiv.org > cs > arXiv:2111.12159, Nov 2021
In this work, we present a music-driven neural framework that generates realistic human motions, which are rich, avoid repetitions, and jointly form a global structure that respects the culture of a specific dance genre. We illustrate examples of various dance genre, where we demonstrate choreography control and editing in a number of applications.
Virtual Dance Museums: the case of Greek/Cypriot folk dancing
Andreas Aristidou, Nefeli Andreou, Loukas Charalambous, Anastasios Yiannakidis, Yiorgos Chrysanthou
Presented at: EUROGRAPHICS Workshop on Graphics and Cultural Heritage, GCH'21, Nov 2021
This paper presentes a virtual dance museum that has been developed to allow for widely educating the public, most specifically the youngest generations, about the story, costumes, music, and history of our dances. The museum is publicly accessible, and also enables motion data reusability, facilitating dance learning applications through gamification.
Emotion Recognition from 3D Motion Capture Data using Deep CNNs
Haris Zacharatos, Christos Gatzoulis, Panayiotis Charalambous, Yiorgos Chrysanthou
Presented at: 3rd IEEE Conference on Games, Aug 2021
Background segmentation in multicolored illumination environments
Nikolas Ladas, Paris Kaimakis, Yiorgos Chrysanthou
The Visual Computer, 37, 2221–2233, Aug 2021
We present an algorithm for the segmentation of images into background and foreground regions. The proposed algorithm utilizes a physically based formulation of scene appearance which explicitly models the formation of shadows originating from color light sources. This formulation enables a probabilistic model to distinguish between shadows and foreground objects in challenging images.
A 3D digitisation workflow for architecture-specific annotation of built heritage
Marissia Deligiorgi, Maria I.Maslioukova, Melinos Averkiou, Andreas C.Andreou, Pratheba Selvaraju, Evangelos Kalogerakis, Gustavo Patow, Yiorgos Chrysanthou, George Artopoulos
Journal of Archaeological Science: Reports, Volume 37, June 2021, 102787, Jun 2021
Adult2Child: Motion Style Transfer using CycleGANs
Yuzhu Dong, Andreas Aristidou, Ariel Shamir, Moshe Mahler, Eakta Jain
Presented at: ACM SIGGRAPH Conference on Motion, Interaction, and Games, MIG'20, Oct 2020
This paper presents an effective style translation method that tranfers adult motion capture data to the style of child motion using CycleGANs. Our method allows training on unpaired data using a relatively small number of sequences of child and adult motions that are not required to be temporally aligned. We have also captured high quality adult2child 3D motion capture data that are publicly available for future studies.
Using Epistemic Game Development to Teach Software Development Skills
Christos Gatzoulis, Andreas S. Andreou, Panagiotis Zaharias, Yiorgos Chrysanthou
International Journal of Game-Based Learning (IJGBL), 10(4), Oct 2020
MotioNet: 3D Human Motion Reconstruction from Monocular Video with Skeleton Consistency
Mingyi Shi, Kfir Aberman, Andreas Aristidou, Taku Komura, Dani Lischinski, Daniel Cohen-Or, Baoquan Chen
ACM Transaction on Graphics, 40(1), Article 1, Sep 2020
Presented at: SIGGRAPH Asia 2020
MotioNet is a deep neural network that directly reconstructs the motion of a 3D human skeleton from monocular video. It decomposes sequences of 2D joint positions into two separate attributes: a single, symmetric, skeleton, encoded by bone lengths, and a sequence of 3D joint rotations associated with global root positions and foot contact labels. We show that enforcing a single consistent skeleton along with temporally coherent joint rotations constrains the solution space, leading to a more robust handling of self-occlusions and depth ambiguities.
Salsa dance learning evaluation and motion analysis in gamified virtual reality environment
Simon Senecal, Niels A. Nijdam, Andreas Aristidou, Nadia Magnenat-Thalmann
Multimedia Tools and Applications, 79 (33-34): 24621-24643, Sep 2020
We propose an interactive learning application in the form of a virtual reality game, that aims to help users to improve their salsa dancing skills. The application consists of three components, a virtual partner with interactive control to dance with, visual and haptic feedback, and a game mechanic with dance tasks. Learning is evaluated and analyzed using Musical Motion Features and the Laban Motion Analysis system, prior and after training, showing convergence of the profile of non-dancer toward the profile of regular dancers, which validates the learning process.
Digital Dance Ethnography: Organizing Large Dance Collections
Andreas Aristidou, Ariel Shamir, Yiorgos Chrysanthou
ACM Journal on Computing and Cultural Heritage, 12(4), Article 29, Nov 2019
This paper presents a method for contextually motion analysis that organizes dance data semantically, to form the first digital dance ethnography. The method is capable of exploiting the contextual correlation between dances, and distinguishing fine-grained difference between semantically similar motions. It illustrates a number of different organization trees, and portray the chronological and geographical evolution of dances.
Why did the human cross the Road?
Panayiotis Charalambous, Yiorgos Chrysanthou
Presented at: ACM SIGGRAPH Conference on Motion, Interaction and Games (MIG ’19), Article 47, 1–2. Best Poster Award, Oct 2019
Humans at rest tend to stay at rest. Humans in motion tend to cross the road – Isaac Newton.” Even though this response is meant to be a joke to indicate the answer is quite obvious, this important feature of real world crowds is rarely considered in simulations. Answering this question involves several things such as how agents balance between reaching goals, avoid collisions with heterogeneous entities and how the environment is being modeled. As part of a preliminary study, we introduce a reinforcement learning framework to train pedestrians to cross streets with bidirectional traffic. Our initial results indicate that by using a very simple goal centric representation of agent state and a simple reward function, we can simulate interesting behaviors such as pedestrians crossing the road through crossings or waiting for cars to pass.
Real-time 3D Human Pose and Motion Reconstruction from Monocular RGB Videos
Anastasios Yiannakides, Andreas Aristidou, Yiorgos Chrysanthou
Comp. Animation & Virtual Worlds, 30(3-4), May 2019
Presented at: Computer Animation and Social Agents - CASA'19
In this paper, we present a method that reconstructs articulated human motion, taken from a monocular RGB camera. Our method fits 2D deep estimated poses of multiple characters, with the 2D multi-view joint projections of 3D motion data, to retrieve the 3D body pose of the tracked character. By taking into consideration the temporal consistency of motion, it generates natural and smooth animations, in real-time, without bone length violations.
