XRF55 S123-Dev / S4-Blind checkpoints
This repository is the durable checkpoint and provenance archive for our 55-class WiFi CSI human-activity-recognition experiments on XRF55.
The current protocol fits model weights on Scene 1, selects development choices with an equal-weight score over a fixed Scene-1 subject holdout and Scenes 2--3, and evaluates frozen models on Scene 4. Scene-4 subjects 03, 04, and 13 are excluded from every Scene-1 training and development role.
The historical/ namespace preserves checkpoints from the earlier
Scene-1-repetitions-01--14 Group-CV protocol. Historical and current results are
not directly interchangeable.
License and intended use
XRF55 is released by its authors under CC BY-NC 4.0. To avoid granting broader rights than the training data permit, checkpoint bundles in this archive are also distributed for non-commercial research use under CC BY-NC 4.0. Project code is separately licensed under Apache-2.0.
No authentication material, private human recordings, video, names, contact information, or unreleased sensor data are included.
Citation
@article{wang2024xrf55,
title={XRF55: A Radio Frequency Dataset for Human Indoor Action Analysis},
author={Wang, Fei and Lv, Yizhe and Zhu, Mengdie and Ding, Han and Han, Jinsong},
journal={Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies},
volume={8},
number={1},
year={2024},
doi={10.1145/3643543}
}
Every published bundle contains a SHA-256 content manifest and a recovery
receipt. Use a pinned repository revision rather than main for scientific
replay.