README.md

metadata

pretty_name: SpecFormer Full-Spectrum Optical Thin-Film Dataset
license: cc-by-4.0
task_categories:
  - other
task_ids:
  - regression
tags:
  - photonics
  - optics
  - materials-science
  - inverse-design
  - thin-films
  - multilayer
  - spectrum
  - parquet
size_categories:
  - 10M<n<100M
configs:
  - config_name: default
    data_files:
      - split: train
        path: train/*.parquet
      - split: validation
        path: val/*.parquet
      - split: test
        path: test/*.parquet

This dataset is the Hugging Face friendly parquet export of the SpecFormer full-spectrum optical multilayer thin-film dataset. It contains 10,000,000 training samples, 500,000 validation samples, and 500,000 test samples. Each sample stores compact uint16 spectral targets together with padded structural labels and fabrication-related constraints for inverse design and spectrum-conditioned modeling tasks.

The 774-dimensional target is ordered as VIS-R, VIS-T, NIR-R, NIR-T, and normal-incidence IR emissivity. Spectral values are stored as uint16 values scaled by 1000; divide by 1000 to recover normalized floating-point spectra in [0, 1].