This dataset contains the database of image snippets of dense clusters of PIV tracer particles (i.e. Mie scattering) marking the reactants within sooting flame measurements in an RQL-type combustors. The database is one subclass utilised in the generation of synthetic training data via domain randomisation. The resulting dataset is subsequently used for training of DL-based image segmentation models. The resulting dataset is subsequently used for training of DL-based image segmentation models: B. Jose, K. P. Geigle, F. Hampp, Domain-Randomised Instance-Segmentation Benchmark for Soot in PIV Images, submitted to Machine Learning: Science and Technology (2025)