The dataset contains the spectrogram images of the real and generated human breathing sounds. The real breathing sounds are procured from Covid-19 patients as well as people not affected by Covid-19 (normal). The generated breathing sounds are procured from online medical repositories for educational purposes. The breathing sounds are segregated into 4 categories: Crackles-coarse, Crackles-fine, Wheezes, and Normal. The dataset comprises spectrogram images for - 420 coarse crackles (generated), - 375 fine crackles (generated), - 1143 normal (generated), - 322 wheezes (generated), - 1734 Covid-19 (real), - 2690 non Covid-19 (real)