Here we have performed meta learning on environmental sounds, available in the ESC-10 and ESC-50 datasets, to check how these algorithms perform on different scenarios, rather than the traditional testing done on images, tabulated the results for different n-way and k-shot settings, helpful for further research in this field. The general trend observed and results are: Training accuracy found to be very good, around 98-99.5% Testing accuracy was highest as expected for higher shots.
All the corresponding code files and test results are uploaded in this repository.