Skip to content

Uncertainty models for adversarial robustness in small-scale hybrid speech recognition

License

Notifications You must be signed in to change notification settings

rub-ksv/uncertaintyASR

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

19 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

uncertaintyASR

build docker

run the following command:

$ docker build -t uncertainty_docker .

run docker

$ docker run --runtime=nvidia -e NVIDIA_VISIBLE_DEVICES=0 \
    --rm \
    -v <path-to-repo>/src:/root/asr-python/src \
    -v <path-to-repo>/exp:/root/asr-python/exp \
    -v <path-to-repo>/results:/root/asr-python/results \
    -v <path-to-dataset>/TIDIGITS-ASE:/root/asr-python/TIDIGITS-ASE \
    -it uncertainty_docker \
    python3 /root/asr-python/src/recognizer_torch.py 'NN'

Depending on the model use 'NN', 'dropout', 'BNN2', or 'ensemble'

must at least contain the wav files for which we want to create adverarial examples.

run eval

After calculating the adversarial examples, the evaluation on the uncertainty features can be called via:

 docker run --runtime=nvidia -e NVIDIA_VISIBLE_DEVICES=0 \
        --rm \
        -v <path-to-repo>/src:/root/asr-python/src \
        -v <path-to-repo>/exp:/root/asr-python/exp \
        -v <path-to-repo>/results:/root/asr-python/results \
        -v <path-to-dataset>/TIDIGITS-ASE:/root/asr-python/TIDIGITS-ASE \
        -it uncertainty_docker \
        python3 /root/asr-python/src/eval.py

About

Uncertainty models for adversarial robustness in small-scale hybrid speech recognition

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published