DeepSpeech model for Swedish ASR
Our fork of Mozilla's DeepSpeech can be found in the v0.6.1-thesis branch of this repository. Notably, we have added a flag to help with transfer learning. It allows taking a checkpoint with x output nodes to output a new checkpoint with y output nodes. This enables transfering from one alphabet to another.
Our Swedish ASR model can be found in the Releases page of this repository.