HypercubeAIData

September 1, 2026 ยท View on GitHub

Datasets for the HypercubeAI family of projects. This repository holds no code; the data lives in the Releases as zip files.

Install

Download the release zip and unpack it into C:\HypercubeAI\data, so the dataset folder lands at:

C:\HypercubeAI\data\RamanSpectraLCOHard\

HypercubeAI projects read the data from that path in place.

RamanSpectraLCOHard

Raman spectra with ground-truth baselines, for training and evaluating baseline-extraction models. About 75 MB zipped, roughly 1 GB unpacked.

RamanSpectraLCOHard/
  Training/      10000 samples, indices 0..9999
  Validation/    2000 samples, indices 0..1999

Each sample index N has three files, each one line of 2048 comma-separated floats:

FileContents
N.data.txtThe spectrum: 2048 amplitudes in counts
N.label.txtThe ground-truth baseline, same length and units
N.peaks.txtThe peak component of the spectrum
xaxis.txtOne per split: the wavenumber axis for all spectra

Treat released zips as immutable; corrections ship as a new release version, never by replacing an existing asset.