Hybrid Image Classification Model using ResNet101 and VGG16

-, G.Surekha; Keerthana, Patlolla Sai; Varma, Niraj; Gopi, T.

doi:10.1109/icaaic56838.2023.10140790

Cited by 1 publication

(1 citation statement)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This model can extract intricate features from facial images through multiple convolutional layers, capturing hierarchical patterns crucial for accurate diagnosis. The method utilizes the VGG16 architecture's ability to learn complex representations [35], allowing it to discern subtle facial features indicative of Down Syndrome, ultimately enhancing the precision and reliability of the diagnostic process. The VGG16 method for the diagnosis of Down Syndrome in Children Using facial images can be represented mathematically as:…”

Section: Convolutional Neural Networkmentioning

confidence: 99%

Novel Transfer Learning Based Deep Features for Diagnosis of Down Syndrome in Children Using Facial Images

Raza,

Munir,

Almutairi

et al. 2024

IEEE Access

View full text Add to dashboard Cite

Down syndrome is a chromosomal condition characterized by the existence of an additional copy of chromosome 21. This genetic anomaly leads to a range of developmental challenges and distinct physical characteristics in affected children. Children with Down syndrome often exhibit specific craniofacial proportions, such as a relatively shorter midface and broader facial width. These distinct facial features, including a flat nasal bridge, almond-shaped eyes, and a small and somewhat flattened head, can serve as valuable indicators for early diagnosis and intervention. This study aims at the early diagnosis of Down syndrome using an advanced neural network approach. We used 3,009 facial images of children with Down syndrome and healthy children taken from the age group range of 0 to 15 for conducting our research experiments. We proposed a novel transfer learning-based feature generation named VNL-Net, which is an ensemble of VGG16, Non-Negative Matrix Factorization (NMF), and Light Gradient Boosting Machine (LGBM) methods. This unique VNL-Net feature extraction initially extracts spatial features from input image data. Then, the ensemble feature set of NMF and LGBM is extracted from spatial features. We built several advanced artificial intelligence-based approaches on the newly created feature set to evaluate performance. Extensive research experimental results show that the logistic regression method outperformed state-of-the-art studies with a high-performance accuracy of 0.99. We also fine-tuned each applied method and validated performance using the k-fold cross-validation mechanism. The runtime computational complexity of the applied methods is also determined. Our proposed innovative research has the ability to revolutionize the early diagnosis of Down syndrome in children using facial images.

show abstract

Section: Convolutional Neural Networkmentioning

confidence: 99%