This work studies deep learning-based approaches to develop systems for acoustic-based train arrival detection (A-TAD). Firstly, we develop a novel audio dataset of train horns, railway alarms, ...