Audio Cats and Dogs

Description

The Audio Cats and Dogs dataset contains 277 barking and meowing audio files from cats and dogs.

Neural network

This is the test with the neural network architecture used to obtain up to –% accuracy on this dataset. :warning: To reach this accuracy you may need more attempts and more learning time.

TEST_F(AudioCatsAndDogsTest, trainNeuralNetwork)
{
    StraightforwardNeuralNetwork neuralNetwork({
        Input(sizeOfOneData),
        LocallyConnected(1, 1000, activation::tanh),
        GruLayer(20),
        GruLayer(5),
        FullyConnected(2)
    });
    neuralNetwork.optimizer.learningRate = 0.002f;
    neuralNetwork.optimizer.momentum = 0.2f;
    neuralNetwork.train(*data, 100_ep || 0.6_acc || 30_s);
    auto recall = neuralNetwork.getWeightedClusteringRate();
    auto accuracy = neuralNetwork.getGlobalClusteringRate();
    ASSERT_RECALL(recall, 0.50);
    ASSERT_ACCURACY(accuracy, 0.6);
}

See the code
This dataset is not executed during the iteration tests because no stable architecture enough accurate was found.