Classification clean/noisy audio

Given mel-frequency spectogram predict whether audio is clean or noisy. Model was trained on training data and evaluated on validation data. Task was solved by simple Keras model. Architecutre is below. To only see the notebook, please use nbviewer