AI Belha Classifier

YAMNet-based audio classification model — static page for the NOSInovacao Hugging Face organisation.

About

This Space hosts the AI Belha Classifier, an audio classification model built on YAMNet. It classifies audio clips into sound categories using a pre-trained YAMNet feature extractor fine-tuned on the AI Belha dataset.

Repository contents

Model files

File Size Description
model/model.h5 ~12.7 MB Keras model weights (classifier head)
model/model.npy ~640 B NumPy array (normalisation / label statistics)
yamnet/yamnet.h5 ~15.3 MB YAMNet feature-extractor model weights

Code

File Description
app.py Gradio application entry point
inference.py Audio inference / prediction logic
features.py Audio feature extraction (YAMNet embeddings)
yamnet.py YAMNet model definition
dataset_preprocess.py Dataset preprocessing utilities
params.py Hyperparameters and configuration
train.py Training script
yamnet_test.py Tests for YAMNet component
requirements.txt Python dependencies

Run locally

Gradio builds are currently failing on Hugging Face infrastructure, but you can run the classifier locally:

# Clone this Space
git clone https://huggingface.co/spaces/NOSInovacao/AI-Belha-Classifier
cd AI-Belha-Classifier

# Install dependencies
pip install -r requirements.txt

# Run the Gradio app
python app.py

Then open http://127.0.0.1:7860 in your browser.

Links

Notes