 This paper investigates the performance of three supervised deep learning methods for automated USV segmentation, an autoencoder neural network, AE, a UNET neural network, UNET, and a recurrent neural network, INN. These models use spectrogram data from recordings of USVs to detect and classify the calls. The authors evaluated their performance using a dataset of manually annotated USVs, and compared it against other state-of-the-art methods. The results showed that all three models achieved high accuracy, with UNET, and AE achieving values above 23008595 backslash percent 23008595 percent outperforming other methods. Furthermore, the authors tested the models on an external dataset, and found that UNET performed best. This suggests that these models could serve as a useful benchmark for future research. We are article.tv, links in the description below.