In recent years, NASA’s Kepler and Transiting Exoplanet Survey Satellite (TESS) missions collected a huge amount of data for identifying transiting planets around stars. Indeed, nearly all stars are now expected to host planets, and as our technology improves we keep finding more of them. However, such an unprecedented wealthy of data also requires new methodologies, based on automatic and data-driven techniques. In this talk, I will present a new machine learning approach, based on a Deep Neural Network, to perform an automatic, efficient and reliable binary classification of TESS light-curves, in planet candidate (PC) and non-transiting planet (NTP), without the needs of human supervision. Firstly the neural network model is learned by using the Kepler data set and secondly a transfer learning mechanism is adopted for classifying TESS light-curves. Moreover, we further expanded the training set obtained from the Autovetter Planet Candidate Catalog hosted at the NASA Exoplanet Archive by the means of augmentation techniques. In this way the model is able to reach a better accuracy than a randomly initialized network. The obtained model is then adopted to classify thousands of TESS light-curves, as extracted by Eleanor. Despite the volume of these data (> 10 Tb), we will be able to process all light-curves within a few weeks, by using an High Performance Computing platform directly accessible to us. Finally, we plan to compare our new PC catalog with the ones produced by other vetting algorithms (e.g. DAVE) in order to assess the accuracy of the model.