README.md

## Model Overview

This model is a Morse Code recognition model. It was trained with the package at https://github.com/1-800-BAD-CODE/MorseCodeToolkit.

This model accepts as input audio signals sampled at 8khz containing Morse code. The model produces the English transcription of the Morse code signal.

For inference, only the base NeMo package needs to be installed because this is just an ASR model trained to decode Morse code signals rather than speech signals.

## How to Use this Model
With NeMo is installed, this model can be used to run inference on Morse code audio files.

### Automatically instantiate the model

```python
import nemo.collections.asr as nemo_asr

asr_model = nemo_asr.models.ASRModel.from_pretrained("1-800-BAD-CODE/morsecode_en_quartznet_10x5")
```

### Transcribing using Python
First, let's download an example Morse code audio file from Wikipedia:

```
wget https://upload.wikimedia.org/wikipedia/commons/0/04/Wikipedia-Morse.ogg
```

Then simply do:
```
asr_model.transcribe(['Wikipedia-Morse.ogg'])
['WELCOME TO WIKIPEDIA, THE FREE ENCYCLOPEDIA THAT ANYONE CAN EDIT.']
```

## Limitations
This model was trained on synthetic Morse code data generated by https://github.com/1-800-BAD-CODE/MorseCodeToolkit. Any Morse code generated with parameters outside of the range of the parameters used to generate the training data will not be well recognized by the model.