Description of the datasetThe dataset includes in-situ melt pool signatures (hyperspectral NIR images) during the Directed Energy Deposition of 316L steel for several classes of thermal anomalies. Thermal anomalies were created during the process by varying the scanning speed. Samples were printed on the MiCLAD machine at the Vrije Universiteit Brussel (Belgium). Process and acquisition parameters: Hardware: Machine: MiCLAD (Vrije Universiteit Brussel) Laser: High-YAG BIMO 1064nm, 2.55mm fibre, flat-top Nozzle: Harald-Dickler HighNo 4.0 Process parameters: Laser power: 600 W Scanning speed: 500/700/900/1100/1300 mm/min Powder: 316L 45-105 um Powder flow rate: 3.5 g/m Layer thickness: 0.2 mm Image characteristics: Camera: 3D-One Avior AX-M25NIR Hyperspectral filter layout: 5x5 (25 wavelengths per image) Description of the files CSV dataset (hyperspectral_nir_meltpool_dataset.csv): List of filename, sample, label, time (ms), X and Z position (mm) and local scanning speed (mm/min) for all melt pool signatures. Thermal anomalies are labelled accordingly: 0 : baseline 1 : edge 2 : underheat 3 : strong underheat 4 : overheat 5 : strong overheat Melt pool signatures (hyperspectral_nir_meltpool_images_*.zip): Raw .tif thermal images of the melt pool taken in-situ. The raw images must debayered to retrieve the spectral information, see the Python function and example script. Python debayer function (debayer.py): Debayering function to retrieve the spectral information from the raw images.