Implemntation of speaker verification models like xvector, ecapp-tdnn to train on the barkopedia invidial dog dataset. A full analysis of the vocalization part. Dataset can be downloaded here.
conda create -n dog_verification python=3.10 -y
conda activate dog_verification
pip install uv
uv sync
Train the model:
uv run train.py HYDRA_ARGS
Run inference and generate 3D embedding visualization:
uv run inference.py HYDRA_ARGS