Make Skeleton-based Action Recognition Model Smaller, Faster and Better

Fan Yang, Sakriani Sakti, Yang Wu, Satoshi Nakamura
7/23/2019
cs.CV

Official code: fandulu/DD-Net (279 stars, MIT, last commit 2020-10-25).

Abstract

Although skeleton-based action recognition has achieved great success in recent years, most of the existing methods may suffer from a large model size and slow execution speed. To alleviate this issue, we analyze skeleton sequence properties to propose a Double-feature Double-motion Network (DD-Net) for skeleton-based action recognition. By using a lightweight network structure (i.e., 0.15 million parameters), DD-Net can reach a super fast speed, as 3,500 FPS on one GPU, or, 2,000 FPS on one CPU. By employing robust features, DD-Net achieves the state-of-the-art performance on our experimental datasets: SHREC (i.e., hand actions) and JHMDB (i.e., body actions). Our code will be released with this paper later.

View on arXivView PDF

Code Implementations(1)

fandulu/DD-NetOfficial100%
MIT

A lightweight network for body/hand action recognition

27953Python, Jupyter NotebookJun 30, 20195 years agoMIT
action-recognitionhand-gesture-recognitionjhmdbshrecsimple-tutorial+1 more

Similar Papers

Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition

91% match1 repo1652024

Constructing Stronger and Faster Baselines for Skeleton-based Action Recognition

90% match1 repo252021

Expressive Keypoints for Skeleton-based Action Recognition via Skeleton Transformation

89% match1 repo92024

SkeleTR: Towards Skeleton-based Action Recognition in the Wild

89% match2023

Mask and Compress: Efficient Skeleton-based Action Recognition in Continual Learning

89% match1 repo22024

One-Shot Action Recognition via Multi-Scale Spatial-Temporal Skeleton Matching

89% match2023

Cite this paper

@article{yang2019make,
  title  = {Make Skeleton-based Action Recognition Model Smaller, Faster and Better},
  author = {Fan Yang and Sakriani Sakti and Yang Wu and Satoshi Nakamura},
  year   = {2019},
  eprint = {1907.09658},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1907.09658}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/1907.09658.svg)](https://codeofpaper.com/paper/1907.09658)

Discussion