Summary of Puppet-master: Scaling Interactive Video Generation As a Motion Prior For Part-level Dynamics, by Ruining Li et al.
Puppet-Master: Scaling Interactive Video Generation as a Motion Prior for Part-Level Dynamics
by Ruining Li, Chuanxia Zheng, Christian Rupprecht, Andrea Vedaldi
First submitted to arxiv on: 8 Aug 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
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 Puppet-Master is an interactive video generative model that can serve as a motion prior for part-level dynamics. The model can synthesize videos depicting realistic part-level motion based on given drag interactions. This is achieved by fine-tuning a large-scale pre-trained video diffusion model with a new conditioning architecture to inject the dragging control effectively. The model introduces an all-to-first attention mechanism, which improves generation quality by addressing appearance and background issues in existing models. Puppet-Master is learned from the Objaverse-Animation-HQ dataset of curated part-level motion clips. The model generalizes well to real images across various categories and outperforms existing methods on a real-world benchmark. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Puppet-Master is a new video generation model that can make videos of objects moving in different ways. It uses a special way of looking at the object’s parts to make the motion look realistic. The model is trained on a special dataset of animations and can generate videos based on just one image and some information about how the object should move. It works well even when it hasn’t seen the type of object or movement before. |
Keywords
» Artificial intelligence » Attention » Diffusion model » Fine tuning » Generative model
