This dataset contains the training, validation, and testing datasets for the preprint "Sound absorption estimation of finite porous samples with deep residual learning." The sound fields are generated with a simplified boundary element method (BEM) of a baffled porous layer on a rigid backing using the Delany–Bazley–Miki model. There are 330 thousand simulations, out of which 300 are for training and validation sets (80/20 %), 15 for an interpolation test set, and another 15 for an extrapolation test set. The paper uses the training and validation datasets to train the neural networks. The network's performance against unseen data is assessed with the test sets. More details on using this data can be found in the GitHub repo: https://github.com/eliaszea/finite-absorber-ML.