Applying Whisper to Air Traffic Control

[jlvdoorn]

Applying Whisper to Air Traffic Control ✈️

As part of my Master's Thesis in Aerospace Engineering at the Delft University of Technology, I fine-tuned Whisper (large-v2 and large-v3) on free and public air traffic control (ATC) audio datasets to create an automatic speech recognition model specialized for ATC. My research sets the new state-of-the-art on the commonly-known ATCO2 and ATCOSIM datasets.

This is the first open-source Air Traffic Control ASR model available 🎉. All the code is published in WhisperATC. A report containing all the findings is available here.

The models can be directly downloaded from HuggingFace 🤗.