Core-collapse supernova (CCSN) are one of the sources to emit GWs that are yet to be detected. Furthermore, they represent an interesting candidate for multi-messenger observation, due to their neutrino and electromagnetic emissions. In this chapter we describe and evaluate search and classification methods for gravitational waves from CCSN explosions based on convolutional and recurrent neural network architectures. The proposed approaches use whitened time-series and whitened time-frequency representations as inputs to the deep learning models. We validate the classification accuracy of the models on CCSN waveforms, derived from state-of-the-art hydrodynamical simulations, and instrumental glitches, using both simulated and real interferometer background noise from Virgo, LIGO and Einstein Telescope. We show the robustness of the described methods in distinguishing CCSN and noise transients, and derive accuracies in multi-class waveform model classification in both a single and multi-detector setup.