Learning the spectrogram temporal resolution for audio classification
Based on paper : Anonymous (2023). LEARNING THE SPECTROGRAM TEMPORAL RESOLUTION FOR AUDIO CLASSIFICATION. In Submitted to The Eleventh International Conference on Learning Representations (https://openreview.net/forum?id=HOF3CTk2WH6)
### 1.0 Download the dataset
First, we will use the speechcommands dataset. This dataset contains +105,000 sounds for 35 classes. The audios are short and clear. Here is the link to download the dataset : https://storage.googleapis.com/download.tensorflow.org/data/speech_commands_v0.02.tar.gz
Extract the data in the ./datasets/ folder. Now you have a folder nammed 'speechcommands' inside the folder 'datasets'
You will also need to create some folders : './runs/' './working/'
# Create a new conda environment diffres
conda env create --name diffres --file=env.ymlActivate the environment
conda activate diffres### 1.2 Start training
If you don't want to train, you can directly go to step 1.3 and start testing the model
To train the model, simply run the main.py file as
python3 main.pyIf you want to specify the preserve ratio and the batch size directly in the shell, you can use the following arguments
python3 main.py --ratio 0.5 --batch-size 128If you want to modify other parameters, juste edit the main.py file.