search

RustAudio / deepspeech-rs

Rust bindings for the deepspeech library
Other
296 stars 24 forks source link

Getting bad results #42

Closed jackpeters667 closed 2 years ago

jackpeters667 commented 3 years ago

I'm using 0.9.0 of the model and I have an audio file that counts up to four.

Getting 'yor' as the output. I've tried other audio files as well. But I still haven't managed to match not even a single word.

Did I miss something? I'm using fish shell and I have set the variables:

#.config/fish/config.fish
set -Ux LIBRARY_PATH /home/user/Downloads/ds
set -Ux LD_LIBRARY_PATH /home/user/Downloads/ds

My Downloads/ds folder contains:

deepspeech-0.9.0-models.pbmm
libdeepspeech.so
native_client.amd64.cpu.linux.tar.xz

Upon running (audio of "one two three four"):

cargo run --release --example client /home/user/Downloads/ds ~/cpal/recorded.wav

I get:

TensorFlow: v2.3.0-6-g23ad988
DeepSpeech: v0.9.0-0-g2d04fbe
2021-05-20 22:53:54.751481: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN)to use the following CPU instructions in performance-critical operations:  AVX2 FMA
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
Model initialized in 13.208433ms.
Decoding done in 219.398192ms. Sample length 5.9945s. Running STT.
STT done in 3.908564354s. Real time factor 0.65203
yor

Did I miss something?