This page presents audio demos from the thesis titled "Bird Vocalization Generation using Deep Learning" by Andrii Shevtsov, supervised by Volodymyr Sydorskyi. The work was completed in a fulfillment of the requirements for the MSc program at the Ukrainian Catholic University, Lviv, Ukraine.
On this page, Ground Truth (GT) samples from 12 random selected bird species are compared with generations from both proposed approaches and baseline methods discussed in the thesis. For each method, one sample is selected from three randomly generated outputs. Ground Truth samples are selected entirely at random. Species are chosen randomly across 3 categories of different occurrence frequency (common, uncommon and rare), similarly to the strategy used in the thesis.
100+ training samples present.
50-100 training samples present.
50 or less training samples present.