Explore cutting-edge research on animatable human avatars in this conference talk. Delve into two groundbreaking papers: SNARF, which introduces differentiable forward skinning for animating non-rigid neural implicit shapes, and MetaAvatar, which focuses on learning animatable clothed human models from limited depth images. Gain insights into neural implicit shapes, backward LBS, meta-learning approaches, and fine-tuning techniques. Examine high-level methodologies, training processes, and comparative results. Witness examples and demonstrations that showcase the potential of these innovative techniques in creating more realistic and adaptable human avatars for various applications.
Overview
Syllabus
Intro
Neural Implicit Shapes
NASA
Backward lbs
Training
Results
Summary
Examples
HighLevel Approach
MetaLearning
MetaLearning Example
MetaLearning Approach
MetaLearning Results
Fine Tuning
Comparison
Conclusion
Taught by
Andreas Geiger
Related Courses
-
Parameter Prediction for Unseen Deep Architectures - With First Author Boris Knyazev
-
Andrea Tagliasacchi: Structured Representations for 3D Computer Vision
-
Implicit Neural Scene Representations
-
Learning 3D Reconstruction in Function Space - Long Version
-
Human-Centric 3D Computer Vision for Future AI Systems
-
Enhancing Image Generation Models Using Photogenic Needles in a Haystack - Meta's Emu Architecture
