This dataset contains a curated database of signal-free background patches extracted from OH-PLIF measurements. These patches correspond to reactant-side image regions without OH signal and are retained to preserve the original experimental background characteristics of the measurement system.
The database is utilised in the generation of synthetic training data via physics-informed domain randomisation. During synthetic image generation, randomly selected background patches are assembled into background canvases onto which extracted OH structures are placed, yielding automatically annotated training images for DL-based semantic segmentation models.
For further details and citation, please refer to the submitted paper:
B. Jose, D. Greenblatt, R. P. Lindstedt, A. Breicher, D. Geyer, O. Lammel, F. Hampp, Transferable DL for in-situ validated LIF segmentation, submitted to the Proceedings of the Combustion Institute (2026)
Data Samples:
; Greenblatt, Daniel