This project uses chest X-ray images from tiny datasets with less than a thousand samples to diagnose pulmonary illnesses, including COVID-19 and tuberculosis. One obstacle in the realm of medical diagnostics is the availability of data. To categorize lung diseases, a new transfer learning technique is applied that makes use of the capabilities of deep convolutional neural networks—VGG16 on the ImageNet dataset. Images are included in a special pipeline before they are classified. This strategy’s efficacy is illustrated by comparison with existing frameworks. Remarkably, the research demonstrates that even simpler classifiers, such pre-trained models and shallow neural networks, can outperform complex systems. By comparing the system to publicly available lung datasets, it is positioned as a competitive option. This method achieves accuracy levels comparable to state-of-the-art models, while embracing the advantage of having fewer trainable parameters. Notably, the VGG16-based model demonstrates performance parity with superior systems, effectively striking a balance between computational efficiency and diagnostic accuracy. This work establishes a novel approach that goes beyond data limitations to deliver accessible and dependable lung sickness diagnosis through a framework that is resource efficient.