Respiratory diseases are the leading cause of many hospital admissions and account for a significant portion of fatalities each year in Europe. Long-term respiratory conditions such as asthma affect millions of people globally. A crucial element in diagnosing and monitoring respiratory disease is assessing lung sounds known as respiratory auscultation. Nevertheless, this process can be automated using Deep Learning (DL) techniques to alleviate the strain on healthcare services. This work offers a comparison of various State-Of-The-Art DL models’, namely ConvNeXt, and Vision and Swin transformers for predicting respiratory diseases asthma and COPD and healthy controls from a novel dataset of lung sound recordings represented by melspectrograms. The research concludes that using ConvNeXt in its’ Base configuration outperforms other networks with metrics including accuracy, sensitivity, precision, specificity and F1 score.