This project focuses on automatic drum transcription (ADT). This task is part of the field of music information retrieval and consists of creating a symbolic representation of the notes played by the drums in a music piece. The state-of-the-art methods for ADT are machine learning models trained in a supervised manner, however, the available public datasets are limited either in size or in realism. That is until recently when a new large dataset has been published, effectively increasing the amount of realistic data to multiple orders of magnitude larger (ADTOF: A large dataset of non-synthetic music for automatic drum transcription by Mickaël Zehren, Marco Alunno, and Paolo Bientinesi, conference ISMIR 2021). But to utilize this new source of information, algorithms for ADT have to be adapted. In this project we try different machine learning techniques to exploit this dataset and publish our findings.