kLarity: Dataset of experimental images and trained segmentation model

Project overviewkLarity is an open-source pipeline for automated detection, segmentation, and geometric characterization of gas bubbles in inline endoscopic shadowgraphy images acquired from a pilot-scale stirred tank bioreactor. The dataset supports spatially resolved characterization of bubble size distribution, bubble shape, specific interfacial area, and local gas holdup.

Dataset description

Images_pos 1-6: Raw shadowgraphy images acquired at six axial probe positions within the reactor, covering three xanthan gum concentrations (0.000, 0.125, and 0.250 wt%), four specific power inputs (0.52–2.24 W L⁻¹), and five superficial gas velocities (0.0063–0.013 m s⁻¹)

Klarity_model.pt: Trained YOLOv11x-seg instance segmentation model weights for bubble detection and mask prediction.

output_parquet: Processed results for each probe position, operating condition, and xanthan concentration, including per-bubble geometry, equivalent diameter, specific interfacial area, and gas holdup