Summary of Sweettok: Semantic-aware Spatial-temporal Tokenizer For Compact Video Discretization, by Zhentao Tan et al.
SweetTok: Semantic-Aware Spatial-Temporal Tokenizer for Compact Video Discretization
by Zhentao Tan, Ben Xue, Jian Jia, Junhao Wang, Wencai Ye, Shaoyun Shi, Mingjie Sun, Wenjin Wu, Quan Chen, Peng Jiang
First submitted to arxiv on: 11 Dec 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 A novel video tokenizer called SweetTok is proposed to overcome limitations in current methods for compactly and effectively discretizing videos. Unlike previous approaches that process flattened local visual patches via direct discretization or adaptive query tokenization, SweetTok uses a decoupling framework to compress visual inputs through distinct spatial and temporal queries via DQAE. This allows SweetTok to efficiently compress video tokens while achieving superior fidelity by capturing essential information across spatial and temporal dimensions. The design also includes a motion-enhanced language codebook tailored for spatial and temporal compression to address differences in semantic representation between appearance and motion information. SweetTok improves video reconstruction results by 42.8% w.r.t rFVD on the UCF-101 dataset, and boosts downstream video generation results by 15.1% w.r.t gFVD. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary SweetTok is a new way to turn videos into tokens that can be used in machine learning models. It’s better than other methods because it takes into account both where things are happening in the video (spatial) and when they’re happening (temporal). This helps SweetTok keep more of the important information from the video, which makes it better at reconstructing the original video and generating new videos that look like the old one. It’s also good at recognizing what’s happening in a few seconds of video, even if it’s never seen that exact scene before. |
Keywords
» Artificial intelligence » Machine learning » Tokenization » Tokenizer
