Lightweight HI source finding for next generation radio surveys

Future deep HI surveys will be essential for understanding the nature of galaxies and content Universe. However, large volume these data require distributed automated processing techniques. We introduce LiSA, a set python modules denoising, detection characterization sources in 3D spectral data. LiSA was developed tested on Square Kilometer Array Science Data Challenge 2 dataset, contains pipelines easy domain decomposition parallel execution. algorithms 2D-1D wavelet denoising using starlet transform flexible source finding null-hypothesis testing. These are lightweight portable, needing only few user-defined parameters reflecting resolution also includes two convolutional neural networks to analyze cubes which separate from artifacts predict properties. All components designed as modular possible, allowing users mix match different create their ideal pipeline. demonstrate performance SDC2 is able find 95% with SNR >3 accurately