|
|
@ -2,10 +2,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
Convert Deepspeech 1.8 released model to 2.x.
|
|
|
|
Convert Deepspeech 1.8 released model to 2.x.
|
|
|
|
|
|
|
|
|
|
|
|
## Model
|
|
|
|
## Model source directory
|
|
|
|
* Deepspeech2x
|
|
|
|
* Deepspeech2x
|
|
|
|
|
|
|
|
|
|
|
|
## Exp
|
|
|
|
## Expriment directory
|
|
|
|
* baidu_en8k
|
|
|
|
* baidu_en8k
|
|
|
|
* aishell
|
|
|
|
* aishell
|
|
|
|
* librispeech
|
|
|
|
* librispeech
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
# The released model
|
|
|
|
|
|
|
|
Acoustic Model | Training Data | Hours of Speech | Token-based | CER | WER
|
|
|
|
|
|
|
|
:-------------:| :------------:| :------: | :------: | :-----------------: |
|
|
|
|
|
|
|
|
Ds2 Offline Aishell 1xt2x model| Aishell | 151 h | Char-based | 0.080447 | |
|
|
|
|
|
|
|
|
Ds2 Offline Librispeech 1xt2x model | Librispeech Dataset | 960 h | Word-based | | 0.068548 |
|
|
|
|
|
|
|
|
Ds2 Offline Baidu en8k 1x2x model | Baidu Internal English Dataset | 8628 h |Word-based | | 0.054112 |
|
|
|
|