Summary of Boximator: Generating Rich and Controllable Motions For Video Synthesis, by Jiawei Wang et al.
Boximator: Generating Rich and Controllable Motions for Video Synthesis
by Jiawei Wang, Yuchen Zhang, Jiaxin Zou, Yan Zeng, Guoqiang Wei, Liping Yuan, Hang Li
First submitted to arxiv on: 2 Feb 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 Boximator is a novel approach for fine-grained motion control in video synthesis, introducing two constraint types: hard boxes and soft boxes. These constraints enable users to precisely define object positions, shapes, or motion paths in future frames. Boximator can be used as a plug-in for existing video diffusion models, preserving the base model’s knowledge while training only the control module. To address training challenges, a self-tracking technique is introduced to simplify learning of box-object correlations. Empirically, Boximator achieves state-of-the-art video quality (FVD) scores, improving upon two base models and further enhancing results after incorporating box constraints. Its robust motion controllability is validated by significant increases in the bounding box alignment metric. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine being able to control the movement of objects in a video. That’s what Boximator does! It’s a new way to make videos that lets you decide how things move and change over time. Boximator has two special tools called “hard boxes” and “soft boxes”. These help you specify exactly where and how objects should be placed or moving in future frames of the video. Boximator is easy to use with existing video-making models, and it even gets better results than those models on their own! When people test it out, they prefer the videos made with Boximator over others. |
Keywords
» Artificial intelligence » Alignment » Bounding box » Diffusion » Tracking
