Skip to content

DATA.md

Latest commit

 

History

History
33 lines (24 loc) · 1.22 KB

DATA.md

File metadata and controls

33 lines (24 loc) · 1.22 KB
Raw

Data format

The repository does not redistribute WiSig. The paper pipeline reads the compact pickle representation used by the WiSig preprocessing workflow.

Each pickle must contain a dictionary with these keys:

{
    "tx_list": [...],
    "rx_list": [...],
    "capture_date_list": [...],
    "equalized_list": [0, 1],
    "data": nested_array
}

data[tx_index][rx_index][date_index][equalized_index] is a NumPy array with shape (samples, length, 2). The last dimension stores the real-valued I and Q components. The paper uses equalized samples (equalized=1) and applies per-sample RMS power normalization before feature extraction.

The supplied configurations expect:

  • source date: 2021_03_15
  • held-out capture date: 2021_03_01
  • 12 receiver identifiers
  • 6 transmitters for ManySig
  • the filtered 100-transmitter population for ManyTx

dpr_rffi.data.splits.build_manifest validates the requested transmitter, receiver, date, and sample-count constraints before constructing a protocol. Protocol seeds and transmitter repetitions are fixed in configs/.

Dataset paths may be overridden without changing the configurations:

python scripts/run_protocol.py ... --data /absolute/path/to/ManySig.pkl