ARC-Net

Official implementation of Activity recognition through capsules

Network architectures

The different proposed ARCNet models are provided below.
A preprocessing code for RealWorld is also provided.

• ARCNet for the RealWorld dataset
• ARCNet for the PAMAP2 dataset
• ARCNet-Small for the RealWorld dataset
• ARCNet-Small for the PAMAP2 dataset

Python requirements

• numpy
• PyTorch (1.4.0)
• pytorch-lightning (0.7.5)
• h5py (2.9.0)
• tables (3.6.1)

Reference

@InProceedings{icmla20,
  author       = "H. Damirchi and R. Khorrambakht and H. Taghirad",
  title        = "ARC-Net: Activity Recognition Through Capsules",
  booktitle    = "IEEE Conference on Machine Learning and Applications (ICMLA)",
  month        = "Dec",
  year         = "2020",
  url          = "https://www.icmla-conference.org/icmla20/index.html"
}

Acknowledgments

The CapsNet implementation in this code was based on CapsNet-pytorch.
We sincerely thank the authors of PerceptionNet for sharing their code and processed PAMAP2 data.