Presentation #402.08 in the session “Mining TESS Data with Machine Learning and Other Advanced Methods”.
The large amount of light curves generated monthly from TESS full frame images (FFIs) of newly observed Sectors requires quick and efficient identification of transit-like signals. Currently, the number of planets found and reported by the TESS team in FFIs is limited by the amount of time it takes to identify good planet candidates by human vetters, and the search is done for targets brighter than T = 10.5 mag. We present our ongoing work on Astronet - a machine learning tool we are developing for identifying transit-like and eclipse-like signals in TESS FFIs light curves produced by the MIT quick look pipeline. By improving the efficiency of Astronet, we hope to ease the burden on human vetters and make it possible to search more targets, beyond the current T = 10.5 mag limit, and find many more planet candidates. Our long term goal is to minimize the human involvement in the process of detecting transiting planets in TESS FFI light curves, thereby making the process more uniform. Automated, characterizable vetting pipelines are also essential for creating uniform planet catalogs suitable for occurrence rate studies.