A minimalistic Tensorflow 2.x Keras layer which applies SpecAugment [1] to its input
Try it in Colab
[1] Park, D. S., Chan, W., Zhang, Y., Chiu, C. C., Zoph, B., Cubuk, E. D., & Le, Q. V. (2019). Specaugment: A simple data augmentation method for automatic speech recognition. arXiv preprint arXiv:1904.08779.