Training Deepspeech on Raspberry Pi with Common Voice?

Hello, I would like to use a raspberry pi for offline voice recognition. I am still quite new to this topic. I tried to follow the guide in the docs , but I get this error:

ERROR: Could not find a version that satisfies the requirement ds_ctcdecoder==0.9.3 (from deepspeech-training)
ERROR: No matching distribution found for ds_ctcdecoder==0.9.3

Then I tried to follow the tutorial on github for the raspberry pi. This works well, but when I install DeepSpeech via pip, I cannot call bin/ for the import process (because the file is not found). Any idea how I could train deepspeech with the german de.tar common voice file?

Training or transcribing? The link you posted is about transcribing. Training on a pi for a reasonable dataset would probably take years.

1 Like

In the end I would like to transcribe the spoken Text (continuous offline voice recognition). The accuracy should be quite good. The existing models are english models. I neues this for german language. How could I create or get a model? Is it possible with common voice?

Just search before you post, there are two great German repos. Accuracy quite good is obviously a quite relative term :slight_smile:

1 Like