CrossFi

Article: Zijian Zhao, Tingwei Chen, Zhijie Cai, Xiaoyang Li, Hang Li, Qimei Chen, Guangxu Zhu*, "CrossFi: A Cross Domain Wi-Fi Sensing Framework Based on Siamese Network" (under review, IEEE Internet of Things Journal (IOT))

1. Data

1.1 Dataset

Dynamic Part of WiGesture Dataset

1.2 Data Process

Each sample has a dimension of 2*t*s, representing the channel, height, and width, respectively.

The 2 channels consist of the amplitude and cosine of phase. t is the sample number, and s is the subcarrier number.

You can process the data refer to RS2002/CSI-BERT: Official Repository for The Paper, Finding the Missing Data: A BERT-inspired Approach Against Package Loss in Wireless Sensing (github.com).

2. How to Run

Here, we provide the common parameters for running the code. You can use the --help option to obtain a full list of parameters, such as "head num of multi-attention".

2.1 Full-shot

python full_shot.py --MMD --task <task name> --class_num <class number> --data_path <data path>

If you do not want to use MK-MMD during training, you can omit this parameter. The same applies to the following sections.

2.2 One-shot

You can also use the code in few-shot and point the shot number as 1.

2.2.1 Cross Domain

python one_shot.py --MMD --task <task name> --class_num <class number> --data_path <data path> --test_list <the one-shot class> --score distance

2.2.2 Cross Class

python cross_class-one_shot.py --task <task name> --class_num <class number> --data_path <data path> --test_list <the one-shot class>

2.3 Zero-shot

python zero_shot.py --task <task name> --class_num <class number> --data_path <data path> --test_list <the zero-shot class>

2.4 Few-shot

2.4.1 Cross Domain

Pre-train

python full_shot.py --MMD --task <task name> --class_num <class number> --data_path <data path> --not_full_shot --test_list <the zero-shot class>

Fine-tuine

python few_shot.py --MMD --task <task name> --class_num <class number> --data_path <data path> --test_list <the zero-shot class> --model_path <pre-train model path> --shot_num <k-shot> --score distance

2.4.2 Cross Class

Pre-train

python full_shot.py --MMD --task <task name> --class_num <class number> --data_path <data path> --novel_class --test_list <the zero-shot class>

Fine-tuine

python  cross_class-few_shot.py --task <task name> --class_num <class number> --data_path <data path> --test_list <the zero-shot class> --model_path <pre-train model path> --shot_num <k-shot>

3. Citation

@misc{zhao2024crossficrossdomainwifi,
      title={CrossFi: A Cross Domain Wi-Fi Sensing Framework Based on Siamese Network}, 
      author={Zijian Zhao and Tingwei Chen and Zhijie Cai and Xiaoyang Li and Hang Li and Qimei Chen and Guangxu Zhu},
      year={2024},
      eprint={2408.10919},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2408.10919}, 
}