KI-Glossar
Das vollständige Wörterbuch der Künstlichen Intelligenz
Point Cloud Generation
Process of synthetic creation of 3D point clouds using neural networks to generate xyz coordinate sets with additional attributes like color or intensity.
Signed Distance Functions (SDF)
Mathematical representation of 3D surfaces where each point in space stores its signed distance to the nearest surface, used for implicit generation.
Voxel-based Generation
Discrete 3D volumetric creation method where space is subdivided into voxels (volumetric pixels) to generate dense three-dimensional structures.
Implicit Neural Representations
Continuous representation paradigm of 3D shapes using neural networks to encode geometry through implicit functions rather than explicit meshes.
3D GANs
GAN architectures adapted to generate three-dimensional data using discriminators and generators operating directly in 3D space.
Volumetric Medical Imaging Synthesis
Synthetic generation of volumetric medical data (CT, MRI) to augment training datasets while preserving anatomical features.
Multi-view Stereo Reconstruction
3D reconstruction technique exploiting correspondences between multiple 2D views to infer the three-dimensional geometry of the scene.
Mesh Generation
Algorithmic process creating 3D meshes composed of vertices, edges, and faces to represent complex surfaces from implicit or volumetric data.
Point Cloud Completion
3D reconstruction task aimed at filling missing regions in partial point clouds by predicting the complete geometry of the object.
Procedural Content Generation (PCG)
Algorithmic method creating synthetic 3D data through rules and parameters rather than manual modeling, used for mass generation.
3D Shape Completion
Reconstruction process inferring the complete shape of 3D objects from partial or noisy observations using deep learning models.
Volumetric Rendering
3D visualization technique calculating volume appearance by integrating optical properties along rays traversing discrete 3D space.
3D Data Augmentation
Set of techniques applying geometric and photometric transformations to 3D data to increase diversity and robustness of training datasets.
Surface Reconstruction from Point Clouds
Algorithm converting discrete point clouds into continuous surfaces by interpolating topology between points to create coherent 3D meshes.
3D Style Transfer
Application of neural style transfer techniques to 3D data to modify visual appearance of objects while preserving their geometric structure.
Synthetic Scene Generation
Automatic creation of complex and realistic 3D environments including objects, lighting, and materials for training computer vision systems.