Summary of Real-time Human Action Recognition on Embedded Platforms, by Ruiqi Wang et al.
Real-Time Human Action Recognition on Embedded Platforms
by Ruiqi Wang, Zichen Wang, Peiqi Gao, Mingzhen Li, Jaehwan Jeong, Yihang Xu, Yejin Lee, Carolyn M. Baum, Lisa Tabor Connor, Chenyang Lu
First submitted to arxiv on: 9 Sep 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
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 addresses the challenge of running human action recognition (HAR) in real-time on embedded platforms. Current state-of-the-art pipelines are hindered by excessive delays due to complex computations. The authors identify the Optical Flow (OF) extraction technique as the primary latency bottleneck and explore the tradeoff between accuracy and latency for standard and deep learning-based OF methods. To address this, they design an Integrated Motion Feature Extractor (IMFE), a novel neural network architecture that significantly improves latency while maintaining recognition accuracy. The RT-HARE system, developed on an Nvidia Jetson Xavier NX platform, achieves real-time HAR at 30 frames per second with high accuracy. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about making it possible to recognize human actions in videos in real-time on devices like smartphones or tablets. Right now, this task takes too long because of the complex calculations involved. The researchers found that a key part of these calculations – called Optical Flow – was slowing things down. They experimented with different ways to do Optical Flow and designed a new way to make it faster and more efficient. This allowed them to create a system that can recognize human actions in real-time, while still being accurate. |
Keywords
» Artificial intelligence » Deep learning » Neural network » Optical flow
