• A.Vanaja
• I.Kullayamma

Vitamin deficiencies can produce visible changes in body regions such as the skin, eyes, lips, nails, and tongue, providing an opportunity for image-based preliminary screening. This paper proposes a deep-learning-based Vitamin Deficiency Detection System using an Enhanced Convolutional Neural Network (CNN), Vision Transformer (ViT), and Swin Transformer with attention mechanisms. The Enhanced CNN, based on EfficientNet-B3, extracts local spatial and texture features while spatial attention emphasizes informative regions. The Vision Transformer captures global relationships among image patches, whereas the Swin Transformer learns hierarchical representations using shifted-window attention. The prediction scores from the three models are combined through an ensemble score-fusion mechanism and evaluated using a threshold-based decision process. The system classifies the input as healthy or deficiency-related and provides the corresponding vitamin category and severity information. Experimental evaluation demonstrates model scores of 49.0%, 47.5%, and 74.1% average for CNN, ViT, and Swin Transformer, respectively, demonstrating the potential of the proposed hybrid framework for preliminary image-based vitamin deficiency screening

Vitamin Deficiency Detection, Deep Learning, Enhanced CNN, EfficientNet-B3, Vision Transformer (ViT)

"A Hybrid CNN–Vision Transformer Framework with Swin Transformer for Image-Based Vitamin Deficiency Detection", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.13, Issue 9, page no.c889-c901, September-2026, Available :http://www.jetir.org/papers/JETIR2609300.pdf

"A Hybrid CNN–Vision Transformer Framework with Swin Transformer for Image-Based Vitamin Deficiency Detection", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.13, Issue 9, page no. ppc889-c901, September-2026, Available at : http://www.jetir.org/papers/JETIR2609300.pdf