v0.1.0-alpha

Initial alpha release of CARSBench.

CARSBench is a BCARS/CARS simulation and domain-generalization benchmark for Raman-retrieval models.

Added

• Frequency-domain CARS/BCARS simulation workflow

• Eight benchmark domain presets:

  • A_typical
  • B_high_res
  • C_low_res_noisy
  • D_calibration_shift
  • E_window_shift
  • F_nrb_family_shift
  • G_biochemical_source
  • H_biochemical_target

• Raman-equivalent target generation

• Chunked .npz dataset writing

• Per-sample metadata export

• Dataset manifest files

• Multi-seed generation workflow

• QC scripts for general and domain-specific checks

• Visualization scripts

• Basic API and reproducibility tests

• GitHub Actions CI for tests, formatting, and linting

• Documentation for domains, dataset format, reproducibility, QC outputs, and results

Notes

• Generated datasets are not included in the repository.
• Full benchmark datasets should be generated locally or stored externally.
• This is an alpha-stage research and portfolio project.
• This project is not intended for clinical diagnosis, medical decision-making, or deployment in real healthcare settings.