{
    "name": "genre Dortmund",
    "type": "multi-class classifier",
    "link": "https://essentia.upf.edu/models/classifiers/genre_dortmund/genre_dortmund-vggish-audioset-1.pb",
    "version": "1",
    "description": "classification of music by genre",
    "author": "Pablo Alonso",
    "email": "pablo.alonso@upf.edu",
    "release_date": "2020-03-31",
    "framework": "tensorflow",
    "framework_version": "1.15.0",
    "classes": [
        "alternative",
        "blues",
        "electronic",
        "folkcountr",
        "funksoulrnb",
        "jazz",
        "pop",
        "raphiphop",
        "rock"
    ],
    "model_types": [
        "frozen_model"
    ],
    "dataset": {
        "name": "Music Audio Benchmark Data Set by TU Dortmund University",
        "citation": "@inproceedings{homburg2005benchmark,\n  title={A Benchmark Dataset for Audio Classification and Clustering.},\n  author={Homburg, Helge and Mierswa, Ingo and M{\\\"o}ller, B{\\\"u}lent and Morik, Katharina and Wurst, Michael},\n  booktitle={ISMIR},\n  volume={2005},\n  pages={528--31},\n  year={2005}\n}",
        "size": "1820 track excerpts, 46-490 per genre",
        "metrics": {
            "5-fold_cross_validation_normalized_accuracy": 0.52
        }
    },
    "schema": {
        "inputs": [
            {
                "name": "model/Placeholder",
                "type": "float",
                "shape": [
                    96,
                    64
                ]
            }
        ],
        "outputs": [
            {
                "name": "model/Sigmoid",
                "type": "float",
                "shape": [
                    1,
                    9
                ],
                "op": "Sigmoid",
                "output_purpose": "predictions"
            },
            {
                "name": "model/logits/BiasAdd",
                "type": "float",
                "shape": [
                    1,
                    9
                ],
                "op": "fully connected",
                "description": "logits",
                "output_purpose": ""
            },
            {
                "name": "model/fully_connected/BiasAdd",
                "type": "float",
                "shape": [
                    1,
                    100
                ],
                "op": "fully connected",
                "description": "penultimate layer",
                "output_purpose": ""
            },
            {
                "name": "model/vggish/fc2/BiasAdd",
                "type": "float",
                "shape": [
                    1,
                    128
                ],
                "op": "fully connected",
                "output_purpose": "embeddings"
            }
        ]
    },
    "citation": "@inproceedings{alonso2020tensorflow,\n  title={Tensorflow Audio Models in Essentia},\n  author={Alonso-Jim{\\'e}nez, Pablo and Bogdanov, Dmitry and Pons, Jordi and Serra, Xavier},\n  booktitle={IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},\n  year={2020}\n}",
    "inference": {
        "sample_rate": 16000,
        "algorithm": "TensorflowPredictVGGish"
    }
}