Summary of Malt: Multi-scale Action Learning Transformer For Online Action Detection, by Zhipeng Yang et al.
MALT: Multi-scale Action Learning Transformer for Online Action Detection
by Zhipeng Yang, Ruoyu Wang, Yang Tan, Liping Xie
First submitted to arxiv on: 31 May 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 This paper proposes a multi-scale action learning transformer (MALT) to tackle online action detection (OAD). MALT combines a hierarchical encoder with multiple encoding branches, allowing it to capture features at various scales. The output from each branch is incrementally fed into the subsequent one through cross-attention calculation, enabling a transition from coarse to fine-grained features as the branches deepen. Additionally, an explicit frame scoring mechanism employing sparse attention filters out irrelevant frames efficiently without requiring an additional network. MALT achieves state-of-the-art performance on two benchmark datasets (THUMOS’14 and TVSeries), outperforming existing models. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper creates a new way to detect actions in videos. It’s like trying to recognize what people are doing from a live stream, without knowing what will happen next. The problem is that actions can be big or small, so the computer needs to understand them at different levels of detail. The researchers created something called MALT (Multi-Scale Action Learning Transformer) to solve this issue. It’s like having multiple pairs of glasses to see things from different perspectives. They also came up with a new way to look at each frame and decide which ones are important, without needing extra help. This new method works better than others on two famous datasets. |
Keywords
» Artificial intelligence » Attention » Cross attention » Encoder » Transformer
