Neural network project addresses the problem of general-purpose automatic audio tagging. (http://dcase.community/challenge2018/task-general-purpose-audio-tagging). This project uses a convolutional neural network (VGG) to classify 41 classes of audio. Dowload the dataset here. To initialize the workspace use the init_work_space.sh bash script. If you want to train a model:

python3 train_spec.py --model_name "name of the new model"

If you want to try a pretrained model, first lunch the init_work_space.sh script with the -m option and then:

python3 test.py --model_name test9