Description
SignFi is a WiFi CSI dataset for sign-language gesture recognition collected by Ma et al. on commodity Intel 5300 NICs. It pairs CSI traces with American Sign Language (ASL) sign labels and includes both a controlled laboratory subset and a SignFi Home subset captured in a domestic environment, which is widely used as a domain-shift / cross-environment test bed in CSI sensing literature.
Modality / size
- Modality: WiFi CSI from Intel 5300 NIC, 30 subcarriers per antenna pair.
- Subjects / scenarios: multiple signers; lab and home environments (
SignFi Home). - Labels: per-sign class labels (hundreds of ASL signs).
Used by (papers)
- CSI self-supervised pretraining work uses
SignFi Homefor few-shot linear evaluation (2/4/6/12 shots). - Standard benchmark in cross-domain WiFi sensing studies.
Notes
- Public release accompanies the SignFi paper (Ma, Zhou, Wang).
- The
SignFi Homevariant is folded in here as an alias rather than a separate note; the only meaningful split between the two is the environment.