Progress to date

These are among the items which we have been working on as steps toward the RAPID pipeline

  • We have stood up our computing environment on Amazon Web Services (AWS), including databases
  • We have prototyped a set of production image differencing pipeline modules
  • We have been primarily utilizing the NASA Open Universe Rubin-Roman image simulationsimage differencing algorithm comparison
  • We have been testing and evaluating different image differencing algorithms
    • ZOGY (Zackay, Ofek, & Gal-Yam 2016), SFFT (Hu+2022), hybrid cross-convolution+SFFT (Hu & Wang 2024)
    • We are also investigating TRANSLIENT (Springer+2024)
    • We have been exploring machine learning algorithms for transient detection (Sedaghat & Mahabal 2017)
  • We have implemented both AWAICGEN (Masci 2009) and SWarp (Bertin 2010) for image reprojection to a common grid
  • We have implemented code to inject further fake transient sources into the existing image simulations for testing recovery completeness
  • We have implemented a scheme for reference images, which will be constructed for sky tiles based on the Roman sky tessellation
  • We detect sources in the difference images using SourceExtractor (Bertin & Arnouts 1996), as well as tools within photutils (Bradley+2024)
  • Output source lists have been written in/read from Kafka
  • We are working on machine-learning-based real-bogus source discrimination and source classification
  • We are working on construction and formatting of the alert packets
  • Increasingly automated pipeline via Virtual Pipeline Operator
  • We have conducted large-scale processing with our prototype pipeline on the entire OpenUniverse set, which goes from input L2 image to source detection from a difference image!