Summary of Puzzleavatar: Assembling 3d Avatars From Personal Albums, by Yuliang Xiu et al.
PuzzleAvatar: Assembling 3D Avatars from Personal Albums
by Yuliang Xiu, Yufei Ye, Zhen Liu, Dimitrios Tzionas, Michael J. Black
First submitted to arxiv on: 23 May 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); Graphics (cs.GR)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary A novel approach to generating personalized 3D avatars from casual photo collections, known as “Album2Human,” is proposed. The challenge lies in reconstructing a faithful avatar from diverse poses, viewpoints, and occlusion. PuzzleAvatar, a foundational vision-language model fine-tuned on personal OOTD albums, encodes appearance, identity, garments, hairstyles, and accessories into learned tokens. These tokens serve as “puzzle pieces” to assemble a faithful 3D avatar. The model can be customized by interchanging tokens. A new dataset, PuzzleIOI, is introduced, comprising nearly 1K OOTD configurations from 41 subjects. Evaluation shows high reconstruction accuracy, outperforming TeCH and MVDreamBooth, as well as scalability and robustness. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine having a personalized 3D avatar created just for you, based on your everyday photos! That’s what this paper is all about. Right now, there are models that can create avatars for celebrities or fictional characters, but they struggle with creating avatars for regular people. The challenge is that our photo collections often contain diverse poses, viewpoints, and occlusion. This paper introduces a new approach called PuzzleAvatar, which uses a special model to generate a 3D avatar from your personal photos. It’s like solving a puzzle! The model can even be customized by changing certain features. The researchers also created a new dataset with many examples of people in different outfits and hairstyles. |
Keywords
» Artificial intelligence » Language model
