Near-Infrared Spectroscopy and Image Classification of Refuse-Derived Fuels for Cement Production

Near-Infrared Spectroscopy and Image Classification of Refuse-Derived Fuels for Cement Production

Description:This dataset contains near-infrared spectroscopy (NIRS) data and RGBimages of 11,526 manually sorted refuse-derived fuel (RDF) particles. The particles were collected from various cement plants and classified into six categories: paper/cardboard, foils, 3D plastics, rubber, foams, and textiles. The dataset was acquired using an at-line conveyor belt setup, featuring a Basler a2A1920-160ucPRO camera and a Viavi 1700 ES NIRS detector (wavelength range: 908–1676 nm). The dataset is imbalanced, reflecting real-world RDF composition. It is intended for machine learning applications in RDF classification to improve fuel quality control in cement production.

Keywords: RDF classification, near-infrared spectroscopy, image dataset, machine learning, cement production, waste management

Data are saved as split 7zip files (finer_split.7z_001, finer_split.7z_002, ..., finer_split.7z_001). To unpack, download all 7zip files and unpack the finer_split.7z_001 file with 7zip or WinRar under Windows or p7zip under Linux. The database is given as SQLite files (.db). This can, for example, be opened with "DB Browser for SQLite" (https://github.com/sqlitebrowser/sqlitebrowser).

Code.7z holds another archive with Python code to 1) host the database on al local flask API server and 2 ) classify RDF with NIRS and RGB images.