University of Ioannina
Αλγόριθμοι για την αποδοτική κίνηση αρθρωτών και μη αρθρωτών χαρακτήρων
Abstract
dc:descriptionIn computer graphics, animation compression is essential for efficient storage, streamingand reproduction of animated meshes. The goal is to compress an animation thatis given as a sequence of meshes with different vertex coordinates per frame. Compressionis often achieved by technique that introduces pseudo bones and weights thatdescribe the influence of each pseudo bone on the vertices. Previous work has presentedefficient techniques for compression by deriving skinning transformations andweights using clustering of vertices based on geometric features of vertices over time.In this work, we present a novel approach that assigns vertices to bone-influencedclusters and derives weights using deep learning through a training set that consistsof pairs of vertex trajectories (input to the neural network) and the correspondingweights (output of the neural network) drawn from fully rigged animated characters.For deriving even better results, a technique for automatically performing persistentlabeling of the bones is presented, based on deep learning. This technique reduces theapproximation error and facilitates the fusion of two or more compression schemesthat represent the same animation sequence.Additionally, we present a method for improving the approximation error offeredby a compression scheme for semi-articulated characters or for characters with highlydeformable parts. This method analyzes the distribution of high error vertices. Suchvertices are subsequently grouped in clusters using agglomerative clustering by theirgeodesic distance. New bones are then introduced for each cluster. The influence ofthe new bones on the vertices is then adjusted so that the overall approximation error is reduced while the maximum number of weights per vertex remains the same.Tools for animation transfer of one source mesh to a new destination mesh or forcombining the animation of two different characters are very useful for animators. Inthis thesis we present an easy-to-use stand alone, efficient, interactive tool for animationtransfer. Finally, we discuss approaches for blending or synthesis of compressedanimation schemes.
Degree
thesis:*- Grantor dc:publisher
- University of Ioannina
- Year dc:date
- 2022
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Moutafidou, Anastasia
- Μουταφίδου, Αναστασία
Subjects
dc:subject × 15- Γραφικά υπολογιστών
- Μηχανική μάθηση
- Τρισδιάστατοι Χαρακτήρες
- Κίνηση Χαρακτήρων
- Deformable Characters
- 3D animation
- Animation Transfer
- Skinning
- Rigging
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Φυσικές Επιστήμες
- Γραφικά υπολογιστή και Σχεδιασμός με χρήση υπολογιστή
- Computer and Information Sciences
- Natural Sciences
- Computer Graphics and Computer-Aided Design
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/53406