Datasets:
BoilingBench-5 — Human-annotated boiling images
Image-centered boiling benchmark containing RGB frames and human annotations associated with BubbleID and BubbleID-Flow. Use human labels as evaluation references for detection, segmentation, vapor-area fraction, and morphology. Keep machine-generated labels and model checkpoints separately identified; checkpoint licenses may differ from image/annotation rights.
Citations: C. Dunlap et al., “BubbleID: A Deep Learning Framework for Bubble Interface Dynamics Analysis,” Journal of Applied Physics, 136, 014902 (2024), publisher page; and A. Fahim et al., “BubbleID-Flow: Machine-Vision Quantification of Vapor Area Fraction in Subcooled Flow Boiling,” manuscript in preparation.
Full card: BoilingBench-5 documentation.