Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{209277,
        author = {Kondagorri Ruchitha and Dr. S. Jhansi Rani},
        title = {A Comparative Study of ResNet50, VGG16, VGG19, And InceptionV3 Transfer-Learning Architectures for Pneumonia Detection and Classification from Chest X-Ray Images},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {5},
        pages = {1212-1220},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=209277},
        abstract = {Pneumonia remains an important respiratory disease for which rapid and reliable image-based screening can assist clinical decision-making. This study presents a controlled comparative evaluation of four convolutional neural network (CNN) architectures—ResNet50, VGG16, VGG19, and InceptionV3—for binary classification of chest X-ray images into pneumonia-positive and normal categories. All networks are initialized with ImageNet-pre-trained weights and adapted using a common classification head containing convolution, max pooling, dropout, batch normalization, a dense layer, and a sigmoid output. To reduce the effect of limited training data and class imbalance, training-only augmentation using horizontal flipping, rotation, and zoom is employed. A two-stage transfer-learning strategy is used: the convolutional base is first frozen and the newly added classification head is trained, followed by selective unfreezing and fine-tuning of deeper layers at a lower learning rate. The reported test results show accuracies of 85.7%, 86.2%, 88.9%, and 91.3% for VGG16, VGG19, InceptionV3, and ResNet50, respectively. ResNet50 also reports the highest F1-score (91%) and AUC (0.96) in the supplied experimental results. The study further examines preprocessing, augmentation, generalization, computational considerations, and practical deployment issues. The results indicate that architectural design and staged fine-tuning are important factors when transfer learning is applied to relatively small and imbalanced chest X-ray datasets.},
        keywords = {Deep Learning, Transfer Learning, Pneumonia Detection, Chest X-Ray, ResNet50, VGG16, VGG19, InceptionV3, CNN, Medical Image Classification.},
        month = {October},
        }