We are sorry, the benchmarking engine is temporarily not available.

However, in the meantime, you are welcome to download the dataset and prepare your models:

Training data

Test data 

The data are stored in .h5 file format, but we have prepared import scripts for Python/R/MATLAB.


  • change path to your directory
  • set parameter "spectra_count= " to the required number of spectra per sample in the training dataset

A detailed description of the benchmark dataset is available here.

Check out publications related to the contest! The first one is providing insights into the dataset and the second one describes the data-processing strategies of the top 5 participants.

Dataset: Benchmark classification dataset for laser-induced breakdown spectroscopy

Contest: Classification of challenging Laser-Induced Breakdown Spectroscopy soil sample data - EMSLIBS contest