sebastianstarke/AI4Animation
AI4Animation
AI4Animation is a deep learning motion controller and framework for generating biped and quadruped locomotion from motion capture data. It provides a system for training and executing neural networks to drive real-time character animations within Unity.
The project features a neural animation blending system that synthesizes novel movements by mixing unaligned motion sources using learned manifolds. It also includes a real-time inverse kinematics solver for computing joint positions based on target goals and scene geometry.
The toolset covers motion capture data processing, including the import and cleaning of FBX, BVH, GLB, and NPZ formats. It further supports manifold-based motion extraction to learn periodic spatial-temporal structures and sensor-to-motion mapping for translating sparse signals into full-body avatar movements.
Features
- Motion Synthesis Models - Provides a deep learning system to generate biped and quadruped locomotion from motion capture data.
- Neural Model Training - Provides a deep learning system for training neural networks that control the movement and behavior of biped and quadruped characters.
- Locomotion Training - Provides a framework for training neural networks to generate biped and quadruped locomotion from motion capture data.
- Neural Animation Drivers - Executes trained deep learning models within a game engine to generate character joint rotations based on real-time inputs.
- Video and Motion Synthesis - Trains deep learning models to generate biped and quadruped locomotion from raw motion capture data.
- Character Animation Runtimes - Executes trained neural networks within a game engine to drive real-time character animations based on inputs.
- Inverse Kinematics Character Animations - Implements a real-time system to compute joint positions for characters to reach target goals using inverse kinematics.
- Neural Animation Runtimes - Provides a framework for executing trained neural networks to drive real-time character motion within Unity.
- Data Processing Pipelines - Implements a pipeline for importing and cleaning animation data from FBX and BVH formats for neural network training.
- Manifold-Based Motion Extraction - Learns periodic features and spatial-temporal structures from datasets to improve alignment and similarity measures for motion capture.
- Motion Phase Manifold Extraction - Learns periodic features and spatial-temporal structures from motion datasets to improve alignment and similarity measures.
- Inverse Kinematics - Provides a real-time solver for computing joint configurations to reach specific target goals.
- Environment-Adaptive Animations - Generates character movements that adapt to scene geometry for actions like sitting or obstacle avoidance.
- Neural Animation Blending - Synthesizes novel movement combinations by imitating animation layering to mix and blend unaligned motion sources.
- Motion Retargeting - Translates sparse sensor signals into full-body avatar motions in real-time while preserving original motion context.
- Imitative Layering - Synthesizes novel movement combinations by blending unaligned motion sources to simulate traditional animation layer mixing.
- Neural Blending Systems - Synthesizes novel movements by mixing unaligned motion sources using learned manifolds.
Star history
How this analysis was created: This summary and feature list are AI-generated from collected project material and can contain mistakes. Stars, license and language are imported from GitHub. Inclusion does not mean that we have tested or audited this project. Check the source documentation for any feature you depend on. Learn more on our About page.
