25 lines (19 loc) · 624 Bytes

Data Layout

The original wildfire image dataset is not included in this repository.

The project expects the following directories when reproducing the full pipeline:

data/raw/: the unmodified source dataset after extraction
data/splits/: canonical train, validation, and test folders created by scripts/prepare_dataset.py

Expected split layout:

data/splits/
  train/
    fire/
    no_fire/
  val/
    fire/
    no_fire/
  test/
    fire/
    no_fire/

The preparation script also writes data/splits/SPLIT_INFO.json with split counts, seed values, and source file metadata for reproducibility.