Pneumonia Detection from Chest X-Ray Images Using Transfer Learning with EfficientNetB1

This research study shows the use of transfer learning employing the EfficientNetB1 model architecture for the aim of automated categorisation of chest X-ray images to detect pneumonia. The model was trained using a labelled dataset of X-ray pictures whose capacity to better generalise had been improved via image augmentation methods. The general accuracy of 96% with precision at 0.98, recall of 0.91 while having an F1-score of 0.94 for the NORMAL class and precision at 0.95 Recall of 0.99 for the PNEUMONIA class. The model worked with excellent accuracy on the training set, but the validation plots of accuracy and loss identified sharp fluctuations, which is another indicator of overfitting. Possible solutions are to apply stronger regularization techniques and modify the model architecture. In spite of all these issues, the model based on EfficientNetB1 exhibits strong potential for an accurate and reliable pneumonia detection and offers promising aspects for assistance in medical diagnostics. This work underscores how transfer learning may be the key to the eventual success of applying medical image classification and outline areas for further refinement toward generalization and clinical applicability.