Deep Learning-Based Analysis of Pediatric Pneumonia Detection in Children using Fine-tuned NasNetMobile Model

Pediatric pneumonia is a significant cause of morbidity and mortality in children worldwide. Traditional methods for detecting the disease are time-consuming, thereby necessitating automated detection methods. Hence, the paper fine-tunes the light-weight NasNetMobile model architecture using multiple normalization approaches and six recent activation functions like Swish, HardSwish, Mish, etc. to enhance model performance and generalization in the efficient detection of pediatric pneumonia in children. The work also conducts a comparative analysis of six state-of-the-art model CNN model architectures to prove the efficacy of the proposed model. The outcomes show that fine-tuned NasNetMobile model using the Mish activation function, combined with batch normalization and dropout, outperforms other configurations with an F-score of 92%.