Speech Recognition project

The goals of this project are:

• Evaluating the feasability of deploying the DeepSpeech STT (Speech-To-Text) model on a device without GPU accelaration.
• Getting a qualitative intuition behind what goes right/wrong in prediction.
• Getting a quantitative intuition behind what goes right/wrong in prediction (e.g. with Word Error Rate).

A PDF report is available here.