Reproduce “Human 3D Avatar Modeling with Implicit Neural Representation: A Brief Survey” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Mingyang Sun, Dingkang Yang, Dongliang Kou, Yang Jiang, Weihua Shan, Zhe Yan, Lihua Zhang
A comprehensive survey details the application of implicit neural representations (INRs) for 3D human avatar modeling, categorizing techniques across body, hand, and head reconstruction. It highlights how INRs enable continuous, detailed representations from various input data, while also outlining current limitations and promising future research directions in the field.
Key insights
- Implicit Neural Representations (INRs) provide continuous 3D shape representations, allowing capture of complex topologies and fine details at arbitrary resolutions.
- Different INR types (Occupancy Networks, Signed Distance Fields, Neural Radiance Fields) are adapted to specific aspects of avatar modeling, from geometry to view-dependent appearance.
- Strategies like canonical space transformation, part-based modeling, and image-based reconstruction address challenges in representing articulated bodies and reconstructing from limited data.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.